A Multi-Proxy Analysis of Australopithecus anamensis Paleoecology in the Omo-Turkana Basin

by Laurence Dumouchel

B.Sc. in Anthropology, May 2011, University of Montreal M.Sc. in Biological Anthropology, May 2013, University of Montreal M.Phil. in Human Paleobiology, March 2016, The George Washington University

A Dissertation submitted to

The Faculty of The Columbian College of Arts and Sciences of The George Washington University in partial fulfillment of the requirements for the degree of Doctor of Philosophy

August 31, 2018

Dissertation directed by

René Bobe Associate Research Professor of Anthropology

Bernard A. Wood University Professor of Human Origins The Columbian College of Arts and Sciences of The George Washington University certifies that Laurence Dumouchel has passed the Final Examination for the degree of

Doctor of Philosophy as of June 6, 2018. This is the final and approved form of the dissertation.

A Multi-Proxy Analysis of Australopithecus anamensis Paleoecology in the Omo-Turkana Basin

Laurence Dumouchel

Dissertation Research Committee:

René Bobe, Associate Research Professor of Anthropology, Dissertation Co- Director

Bernard A. Wood, University Professor of Human Origins, Dissertation Co- Director

Jonathan G. Wynn, Program Director, National Science Foundation, Department of Earth Sciences, Committee Member

W. Andrew Barr, Visiting Assistant Professor of Anthropology, Committee Member

ii

© Copyright 2018 by Laurence Dumouchel All rights reserved Acknowledgements

I would like to thank my committee, particularly my director Dr. René Bobe who was the first to spark my interest for paleoecology research. I also want to thank my co-director

Dr. Bernard Wood, who graciously stepped in as “official” dissertation director when my original advisor had to leave the George Washington University. A special thank you as well to Dr. W. Andrew Barr, who took on the role of unofficial advisor at times. I am grateful to Dr. Jonathan Wynn for welcoming me to his laboratory in Florida and overseeing my research as it progressed. I thank Dr. Mikael Fortelius and Dr. Laura

Bishop for agreeing to spend time helping me improve this dissertation.

I would like to thank the researchers who currently work at the localities of Allia

Bay, Kanapoi and Mursi, or have worked with specimens from these localities in the past.

These researchers include members of my committee as well as Dr. Fredrick Kyalo

Manthi, Dr. Carol Ward, Dr. J. Michael Plavcan, Dr. Meave Leakey, Dr. Leslea Hlusko,

Dr. Craig Feibel, Dr. Michelle Drapeau and Dr. Denis Geraads, amongst others. I also want to express my gratitude to Rose Nyaboke, Dr. Job Kiibi, Tomas Getachew and all the other museum curators and staff at the Nairobi National Museum, in Kenya and the

National Museum of Ethiopia for their immense help and for the laughs during tea breaks.

I also want to acknowledge the funding sources that supported my research: The

Leakey Foundation, Sigma Xi Grants-in-Aid of Research, Explorers Club Washington

Group Inc., Evolving Earth Foundation, Cosmos Club Foundation and the Lewis N.

Cotlow Fund. I want to extend my appreciation to everyone who wrote me a letter of recommendation over the years, which includes members of my committee and Dr.

Briana Pobiner.

I am grateful to all members of the CASHP community, especially to my co-advisees and friends David Patterson, Amelia Villaseñor and Enquye Negash. I want to recognize the work of Dr. Michelle Drapeau and the members of the anthropology department at the University of Montreal, where I obtained my baccalaureate and my

Master’s degree. I would not be here without the passionate teachers I encountered there.

I also want to thank the members of the “ERG” Ecology Reading Group held at the

Smithsonian National Museum of Natural History and led by Dr. Kay Behrensmeyer for five years of paleoecology-centered discussions. A special thank you to my academically inclined friends and main moral support, Clare Kimock and Emeline Raguin. Thank you to my other friends, my partner, my dog Juniper and to my family, especially to my parents who taught me the importance of both education and dedication. A special mention goes to my dog Lila, who died just a few days before turning in my final thesis draft.

The final “thank you” goes to inspiring mentors from my daily life and to pioneers from the disciplines of paleontology and anthropology, including Mary Anning,

Jane Goodall and others who paved the way for women to pursue scientific research careers.

ii Abstract of Dissertation

A Multi-Proxy Analysis of Australopithecus anamensis Paleoecology in the Omo-Turkana Basin

Australopithecus anamensis, possibly the earliest fully bipedal hominin, lived in eastern

Africa c.4 million years ago (Ma). Three fossil localities in the Omo-Turkana Basin

(Kanapoi, Allia Bay and Mursi) preserve sediments from c.4 Ma. However, the fossil evidence for A. anamensis within the Omo-Turkana Basin is not equally distributed across the three sites. The majority of the fossils within the Omo-Turkana Basin attributed to A. anamensis come from Kanapoi (c.70%), c.30% come from Allia Bay, and none come from Mursi. Preliminary paleoecological analyses suggest that there were differences in the environments of these three sites. This dissertation project tests hypotheses relating hominin abundance to habitat and answer the following overarching question: What were the paleoenvironments of Australopithecus anamensis in the

Omo-Turkana Basin and how did they vary among the three known penecontemporaneous fossil localities? This project uses a multiproxy approach and combines taxonomic, stable isotopic, ecomorphological, mesowear and taphonomic data taken from faunal fossils assemblages from each locality to reconstruct the paleoenvironments.

Chapter 1 introduces the localities and the research objectives. Chapter 2 is an analysis of ungulate mesowear focused on the site of Kanapoi. Chapter 3 is a multi-proxy paleoecological analysis of from Allia Bay. Chapter 4 provides insights into humidity in the Omo-Turkana Basin around 4 Ma by studying taphonomy and the paleoecology of suids. Chapter 5 presents the overarching conclusions of the dissertation.

iii This dissertation allows for a better understanding of the paleoenvironmental context of the earliest obligate biped. Analyses presented here reveal that A. anamensis was more common in relatively open and dry habitats and absent in humid and more closed settings. These conclusions are in line with A. anamensis behavioral and

morphological reconstructions, including C3 hard-object feeding as well as traits characterizing the hominin lineage including an elongated body plan, the loss of hair, the ability to sweat, and obligate bipedal locomotion.

iv Table of Contents

Acknowledgments ...... iv

Abstract of Dissertation ...... vi

List of Figures ...... ix

List of Tables ...... xi

Chapter 1: Introduction ...... 1

Chapter 2: Paleoecological implications of dental mesowear and hypsodonty in fossil ungulates from Kanapoi…………………………………………………………………………………..19

Chapter 3: Multi-proxy analysis of fossil Bovidae (Mammalia, Cetartiodactyla) from the Middle Pliocene Allia Bay, East Turkana, Kenya and implications for Australopithecus anamensis paleoecology………………………………………………42

Chapter 4: Pliocene suids (Mammalia, Cetartiodactyla) from Allia Bay and the paleoenvironments of Australopithecus anamensis in the Omo-Turkana Basin…………………………………………..……….97

Chapter 5: Conclusion ...... 109

References ...... 141

v List of Figures

Chapter 1: Introduction

Figure 1: Geographical map showing the location of Mursi, Allia Bay and Kanapoi within the Omo-Turkana Basin and Africa ...... 7

Figure 2: Stratigraphic sections of Kanapoi, the Southern Allia Bay Plains and Mursi (Yellow Sands) ...... 8

Chapter 2: Paleoecological implications of dental mesowear and hypsodonty in fossil ungulates from Kanapoi

Figure 3: Illustration of traits used in the evaluation of mesowear and hypsodonty…...... 27

Figure 4: Distribution of mesowear scores in the Kanapoi fossils, extant eastern African bovids and other fossil localities ...... 30

Figure 5: Distribution of the mesowear score and hypsodonty index for the 19 Kanapoi bovid specimens for which both measurements are possible ...... 31

Chapter 3: Multi-proxy analysis of fossil Bovidae (Mammalia, Cetartiodactyla) from the Middle Pliocene Allia Bay, East Turkana, Kenya and implications for Australopithecus anamensis paleoecology

Figure 6: A selection of bovid fossils from Allia Bay ...... 57

Figure 7: Comparison of the bovid composition between the three collections, excluding indeterminate specimens ...... 79

vi Figure 8: Stacked bar plots showing the relative proportions of habitat predictions based on astragalar ecomorphology for the assemblages from Allia Bay and Kanapoi ...... 80

Figure 9: Mean hypsodonty against mesowear score by tribe in the three collections ...... 81

Figure 10: δ13C and δ18O values of bovid tribes by fossil collection ...... 88

Chapter 4: Pliocene suids (Mammalia, Cetartiodactyla) from Allia Bay and the paleoenvironments of Australopithecus anamensis in the Omo-Turkana Basin

Figure 11: A selection of suid fossils from Allia Bay ...... 107

Figure 12: Comparison of the suid taxonomic composition between the sites of Allia Bay, Kanapoi and Mursi ...... 125

Figure 13: δ 13C and δ 18O values of Suidae enamel by collection ...... 128

Figure 14: Mosaic plots describing the bone surface modifications from selected tahonomic agents ...... 132

vii List of Tables

Chapter 2: Paleoecological implications of dental mesowear and hypsodonty in fossil ungulates from Kanapoi

Table 1: Mesowear and hypsodonty in Kanapoi ungulates ……………...... 25

Chapter 3: Multi-proxy analysis of fossil Bovidae (Mammalia, Cetartiodactyla) from the Middle Pliocene Allia Bay, East Turkana, Kenya and implications for Australopithecus anamensis paleoecology

Table 2: Measurements (in mm) for all horn cores in the Allia Bay assemblage ...... 53

Table 3: Measurements (in mm) for all Bovidae molars in the Allia Bay assemblage…………………………………………………………………………….....64

Chapter 4: Pliocene suids (Mammalia, Cetartiodactyla) from Allia Bay and the paleoenvironments of Australopithecus anamensis in the Omo- Turkana Basin

Table 4: Measurements (in mm) for all Suidae molars identified to at least the genus level in the Allia Bay assemblage ...... 115

viii Chapter 1: Introduction

Bipedality is widely regarded as a cardinal feature that separates hominins from non-hominins. The fossil record suggests that this behavior appeared in our lineage much earlier than other traits typically associated with hominins, such as tool use and expanded brain size (Harcourt-Smith 2007; Macho 2014). Therefore bipedal locomotion is used as one of the primary criteria to select the taxa that are included in the hominin clade.

According to that definition, scientific consensus places the first australopith species,

Australopithecus anamensis, as the earliest hominin currently known (Wood and

Harrison 2011). Features of the tibia of this species suggest that its knee was habitually directly above the foot; thus this species was possibly the earliest hominin (Leakey et al.

1998; Wood and Leakey 2011).

The species that directly precedes Australopithecus anamensis in the fossil record,

Ardipithecus ramidus (White et al. 1995), is also an early hominin candidate, but its preferred mode of locomotion is debated in the scientific community (Harrison 2010;

Kozma et al. 2018; Wood and Harrison 2011). Similarly, Sahelanthropus tchadensis

(Brunet et al. 2002) and Orrorin tugenensis (Senut et al. 2001) are early hominin candidates, but are currently represented by either contentious or very fragmentary fossil records (Harcourt-Smith 2007; Wood and Harrison 2011).

Despite the importance of A. anamensis for understanding the origins of human bipedality, the species’ environmental context is poorly understood. Thus, the aim of this dissertation is to investigate the paleoecology of Australopithecus anamensis. The majority of A. anamensis fossils, including the species holotype, are from c.4 Ma sites in

1 the Omo-Turkana Basin (Leakey et al. 1995; Wood and Leakey 2011). The fossil localities of Mursi (Ethiopia), Allia Bay (Kenya) and Kanapoi (Kenya) are presently the only sites within the Omo-Turkana Basin to have to the potential to provide data that can be used to reconstruct the paleoenvironmental context of A. anamensis. The site of Fejej, also in the Omo-Turkana Basin, preserves sediments of approximately the same age as

Allia Bay, Kanapoi and Mursi (Kappelman et al. 1996), and some hominin teeth dated to c.4 Ma were recovered from this site, but few faunal remains corresponding to this age were found. Remains attributed to A. anamensis have also been found at sites in the

Awash valley, Ethiopia (White et al. 2006), but by limiting this study to a single sedimentary basin we can reduce the potential influence of differences in geography and geological history (e.g., the rate of sediment supply, the hydrology, the tectonic context)

(Bobe et al. 2007a; Leeder 2009).

1.1 Research Questions

Although A. anamensis probably occurred throughout the Omo-Turkana Basin c.4

Ma, the majority of the fossils attributed to this species have been found at Kanapoi

(c.70%), some have been discovered at Allia Bay (c. 30%) (Ward et al. 2013) and none have been found so far at Mursi. The research question addressed by this dissertation hinges on the relationship between hominin abundance and ecology. What were the paleoenvironments of Australopithecus anamensis in the Omo-Turkana Basin and how did they vary among the three sites? In other words, did the paleoenvironments

2 differ at the two known A. anamensis sites, and did they differ from the paleoenvironment at Mursi?

Fossil hominins are absolutely and relatively rare finds on the paleolandscape, so in order to have large enough sample sizes for effective analysis, we have to study the remains of more abundant penecontemporaneous to shed light on the ecological preferences of hominins (Bobe and Leakey 2009). This dissertation will analyze faunal fossil assemblages from three c.4 Ma localities of Kanapoi, Allia Bay and

Mursi, to better understand the paleohabitats of A. anamensis within the Omo-Turkana

Basin. Comparing paleoecological data across sites or paleontological localities is not a frequent approach in paleontology, especially not in paleoanthropology. The use of this approach is an important contribution of this dissertation.

1.2 Theoretical framework

Most theories that seek to explain the development of human bipedality are rooted in ecological concepts. One of the earliest theories on the origin of bipedality is Dart’s

Savannah Hypothesis. Formulated in 1925 but inspired by Lamarckian ideas from the

19th century (Lamarck 1809), the hypothesis states that human ancestors were forced to become bipeds following the expansion of open environments (Dart 1925). Despite the fact that the hominin fossil record was still poorly known at the time, Dart’s hypothesis influenced many subsequent hypotheses, which also provided ecological explanations for the development of bipedality in early hominins. The Foraging Hypothesis suggests that early hominins adopted bipedal locomotion because they gathered fruits from small trees,

3 reaching for low branches while standing bipedally on the ground (Hunt 1994). The

Thermoregulation Hypothesis posits that bipedality is an adaptation to cope with high solar radiation in open conditions (Wheeler 1991). Finally, the male provisioning model, or Carrying hypothesis, postulates that a male hominin provided for his mate and offspring by carrying food in his arms to bring it back to the home base (Lovejoy 1981).

More recent work has shown just how complex the relationship is between locomotion and habitat. For example, bipedality has been linked to the progressive expansion of open habitats in eastern Africa around 8-5 Ma (Lee-Thorp and Sponheimer

2015) when there was a documented expansion of tropical C4 grasses, which grow in open and unshaded environments (Cerling et al. 1998). The 8-5 Ma period also corresponds to the time during which the first traits of facultative bipedality in hominids like Ardipithecus likely appeared (White et al. 2009b). There is also a more definitive aridification documented in eastern Africa around 3 to 4 Ma, which is hypothesized to be linked to either the onset of glaciation in the Northern Hemisphere (deMenocal 2004), or the closing of the Indonesian seaway (Cane and Molnar 2001).

This evidence suggest that the origin of bipedality in early hominins is somehow connected to the progressive expansion of more open habitats. Given the importance of confidently linking environmental variables to the origin of human bipedality, it is surprising that the environments of the early obligate biped Australopithecus anamensis are so poorly understood. Thus the debate about the origin and success of bipedal locomotion is still largely unresolved (Pawłowski and Nowaczewska 2015).

4 1.3 Context

The International Stratigraphic Chart, International Commission on Stratigraphy (Cohen et al. 2017) defines the Pliocene as the epoch occurring between 2.58 and 5.333 Ma, and it is often described by researchers as a period of transition in Earth’s History. During this epoch, temperatures around the globe transitioned from relatively warm to much cooler and ice sheets started expanding (Dowsett et al. 1999). Atmospheric pCO2 levels are reconstructed as ~380 ppm, which is comparable to modern-day levels, and the global sea level was about 25 meters higher than the present (Dowsett 2007; Seki et al. 2010). In the mid-Pliocene, the flora and fauna start to be similar to present times both in terms of the species represented and their biogeography (Dowsett 2007).

In eastern Africa, the c. 4 Ma period is reconstructed as generally relatively warm, wet, associated with the creation of large lakes and marked by the reduction of deserts

(Brierley et al. 2009; Dowsett 2007). Particularly relevant here is the formation of paleolake Lonyumun, which filled most of the Omo-Turkana Basin at its maximum extension, around 4.1 Ma (Feibel 2011). The paleolake and its modern counterpart, Lake

Turkana, even inspired the name of the species Australopithecus anamensis (‘anam’ means ‘lake’ in Turkana) (Leakey et al. 1995).

1.3.1 History of Research in the Omo-Turkana Basin

The Omo-Turkana Basin is a geological basin preserving fossiliferous sediments from the

Miocene to the Holocene, situated in modern-day southern Ethiopia and northern Kenya

5 and known for its numerous hominin and hominid fossils. Count Samuel Teleki and

Ludwig von Höhnel were the first explorers to formally study the Omo-Turkana Basin area, in 1887-1888 (1976; Mohr 1991). Bourg de Bozas then explored the valley in 1901 and a member of his team, the naturalist Emil Brumpt, collected the first vertebrate fossils (Guillemot 1997). In 1932-1933, the French-led Mission Scientifique de l’Omo explored the lower Omo Valley and various geological and paleontological efforts occur in the years following (Coppens et al. 1976). The first extensive paleontological exploration of the lower Omo valley was led by the International Omo Research

Expedition (IORE) between 1967 and 1976 (Eck, 2007, Coppens et Howell, 1985). IORE was formed by teams from France, the United States and Kenya. While the American and

French teams worked on the Shungura and Usno Formations, the Kenyan team led by

Louis Leakey and his son Richard explored the Mursi and Kibbish Formations. Although

Richard Leakey was aware that the Koobi Fora area was fossiliferous, he “re-discovered” the exposures as they were flying over them. The creation of the Koobi Fora Research

Project (formerly East Rudolf Research Project) shortly followed (Leakey and Leakey

1978). Fieldwork at Kanapoi began in 1966 and 1967 with the Harvard expeditions

(Patterson 1966; Patterson et al. 1970; Patterson and Howells 1967). The excavation of the site is now a central activity of the West Turkana Paleontology Project, codirected by

Dr. Fredrick Kyalo Manthi, Dr. Carol Ward and Dr. J. Michael Plavcan. Allia Bay was excavated by a team from the National Museums of Kenya led by Dr. Meave G. Leakey between 1995 and 1997 (Ward et al. 1999) and paleontological research took place at the

Mursi Formation between 2009 and 2014 (Drapeau et al. 2014).

6

Figure 1: Geographical map showing the location of Mursi, Allia Bay and Kanapoi within the Omo-Turkana Basin and Africa.

7

Figure 2: Stratigraphic sections of Kanapoi, the Southern Allia Bay Plains and Mursi at Yellow Sands. Data from (Brown and Feibel 1991; Drapeau et al. In prep; Feibel 2003)

8 1.3.2 Geology

Kanapoi

Most of the mammalian fossil remains from the Kanapoi assemblage originate from the

4.17-4.07 Ma time interval defined by two tuffs, the Kanapoi tuff and the lower pumiceous tuff (Feibel, 2003; Leakey et al., 1998). Kanapoi fossils were accumulated by the Kerio paleoriver during two distinct fluvial intervals separated by a lacustrine interval

(Leakey et al. 1998). The Kanapoi Formation is located in West Turkana in Kenya. It encompasses sediments dated to 4.195-3.4 Ma, with a Mio-Pliocene Basalt at the base of the sequence and the 3.4 Ma Kalokwanya Basalt at the top of it (Feibel 2003) (Figure 2).

The exposures are 37.3 to 60 m thick and exposed over about 200 km2. The formation is in the dry floodplains in the lower Kerio River valley. The Kerio river is the smallest of three rivers – along with Turkwel and Omo - that feed into the Omo-Turkana Basin. The sedimentary sequence contains fluvial depositions from both before and after the

Lonyumun Lake interval. The paleosols indicate that hominin remains were associated with relatively open semi-arid low tree-shrub savanna habitats (Wynn 2000)

Kanapoi is also where the majority (c.70%) of the remains attributed to A. anamensis in the Omo-Turkana basin have been found (Ward et al., 2013). Several

Australopithecus anamensis fossils originate from this site, including the species holotype, KNM-KP-29281, a mandible and a temporal fragment (Ward et al., 2001).

Other important finds at Kanapoi include a distal humerus discovered in 1965

9 (KNM-KP-271) (Patterson & Howells, 1967) and the tibia fragments used to identify the species as a biped (KNM-KP-29285) (Ward et al., 2001).

Mursi

Due to similarities with the Kanapoi fauna, the fossilized remains found at Mursi are estimated to be about 4.2 Ma (Drapeau et al. 2014). The Mursi Formation is located in southern Ethiopia, and is part of the Omo group. The fossiliferous sediments are overlaid by the Mursi Basalt, dated to approximately 4 Ma (Figure 2). The Mursi Basalt is in fact composed of a series of thin flood basalts, and the error range associated with the radiometric dating is large (Drapeau et al. 2014). Together, the Mursi, Kataboi,

Lothagam, Harr, Kokoi, Usno and Gombe Basalts form the ‘Gombe Group’, a term coined by Haileab (2004) to describe the group of flood basalts accumulated between 4.2 and 3.91 Ma in the Omo-Turkana Basin. The littoral-alluvial depositional settings at the

Mursi Formation are evidence of freshwater lake deposits, seasonal flooding and decomposing vegetation (Drapeau et al. In prep). The paleosols are also indicative of closed and wooded vegetation (Drapeau et al. In prep). At the Mursi Formation, the sedimentary sequences are up to 150 meters thick and exposed over 140 km2 (de

Heinzelin 1983; Drapeau et al. 2014). The fossils were accumulated in a deltaic context and come from two sub-localities: the Yellow Sands, where the fossils have a distinct orange-yellow coloration indicating that they are likely to come from a single geological unit, and Cholo, described in Drapeau and colleague’s (2014) summary article on the

10 formation. For the purpose of this thesis, the fossil collections from the two sub-localities will be combined hereafter referred to as “Mursi”.

Fossil collection efforts at the Mursi Formation have so far yielded about 800 faunal fossils (256 of which are published), but no hominin remains (Drapeau et al.

2014).

Allia Bay

Australopithecus anamensis fossils were recovered from the Lonyumun Member of the

Koobi Fora Formation at Allia Bay, locality 261-1 (3°38' N and 36°16' E), hereafter referred to as the site of “Allia Bay” (Figure 1). Although Allia Bay was recognized as a fossil-bearing locality in early prospections of the Koobi Fora Formation, Feibel (1988) was the first to formally identify and name the locality. He describes the sediments as sandstones containing highly polished fossilized bones and teeth. The Allia Bay fossils were deposited in a fluvial context and represent a single depositional event (Coffing et al. 1994). They were found in a bone bed situated below the Moiti Tuff, dated to

3.97+/- 0.032 Ma (McDougall & Brown, 2008). The bone bed is about 200 m2 in extent and 20 cm thick and is estimated to be approximately 3.98 Ma based on stratigraphic scaling (Coffing et al. 1994; McDougall and Brown 2008) (Figure 2).

The site of Allia Bay is from a time interval slightly younger than those represented by Kanapoi and Mursi. At that point, the Lonyumun Lake was regressing and an extensive fluvial floodplain, called the Moiti floodplain, formed around the meandering Omo paleoriver (Feibel 2011). At that time, sedimentation rate was high and

11 Feibel (2011) suggests that this is also linked to high fossil accumulation around 3.98 Ma at Allia Bay. The low sedimentation rate phase that followed is linked to some of the fossil mammals being poorly preserved.

Comparability of the assemblages

The three assemblages were accumulated in different depositional environments, and therefore some differences, such as the size, the quality and the quantity of fossil remains preserved, are to be expected. Throughout this thesis, an effort was made to control for some of these variables, which have an undeniable effect on the comparability of the assemblages: the size of remains was tested in Chapter 3 and the taphonomic differences between assemblages were thoroughly studied in Chapter 4. Results from all chapters were also appropriately nuanced by previous knowledge of these environments.

In addition, there is a possible age discrepancy between Mursi and the other two assemblages. The only absolute indicator of age in the Mursi sedimentary sequence is the overlying basalt, meaning that the assemblage could in theory be several years older than the other two. However, faunal correlations suggest similar ages for Mursi and Kanapoi

(Drapeau et al., 2014). Scientific work ongoing at Mursi may eventually shed light on this question. In the meantime, this thesis assumes the three assemblages are penecomtemporaneous.

There are also differences between the type of assemblages: Allia Bay is a bone bed and Mursi and Kanapoi are whole formations, thus there is presumably a lot of time averaging represented in the later assemblages. The Mursi fossils come from a single

12 horizon, but it is difficult to estimate how much time that represents. Kanapoi fossils likely show the most time averaging since fossils come from several horizons. However, several age markers are present in the formation, so it is possible to evaluate how much time is represented by each horizon. There are unavoidable differences in the collection strategies used to recover the fossils at Allia Bay, Mursi and Kanapoi, but the strategies are all extremely thorough when it comes to recovering bovid and suid remains, which leads us to conclude that the three collections are broadly comparable on that respect. The site of Allia Bay was excavated and parts of it were sieved (Hagemann 2010). Kanapoi fossils were recovered using multiple strategies, including excavations, targeted surface collections focused on collecting the most diagnostic specimens and systematic description of all surface material, with the fossils left in place. All Mursi specimens were recovered through surface collections, but the team is systematic in its approach and collects all specimens, including non-mammals and diagnostic fossil remains.

1.4 How are the paleoenvironments of Australopithecus anamensis currently described?

The scientific consensus is that A. anamensis lived in a “mosaic” habitat. This is based on the fact that paleoenvironment and vegetation at Kanapoi have been reconstructed as heterogeneous, but relatively open (Geraads et al. 2013; Harris et al.

2003; Ungar et al. 2018; Wynn 2000). The term mosaic suggests a habitat in which both closed (dense canopy) and open (sparse canopy) habitats are present (Behrensmeyer

1985; Behrensmeyer and Reed 2013; Kingston 2007; Reynolds et al. 2011; Verhaegen

13 and Puech 2000). The “mosaic habitat” concept is used to describe several hominin sites

(e.g. Peters and Blumenschine 1995; Su and Harrison 2008). This is problematic, because the term could be applied to a wide range of habitats that differ in important ways, for mosaic environments can incorporate multiple types of habitats (grasslands, bushlands, woodlands, etc.) in different proportions (Reed 1997).

In addition to problems of definition, the claim that A. anamensis lived in a mosaic habitat is mostly based on evidence from the site of Kanapoi; thus specifics regarding the exploitation or the rejection (Mursi) of other environments by A. anamensis need to be further explored. Detailed paleoecological comparisons of A. anamensis sites have the potential to provide meaningful and precise information on the niche of this important taxon.

1.5 The multi-proxy approach in paleoecology

As the field of paleoanthropology progresses, new techniques used to reconstruct ancient environments have emerged and researchers are increasingly aware of the different limitations and biases linked to the results of each of these different methods. To address these challenges, paleoecologists have relied increasingly on the use of multiple methods in combination, i.e., the multi-proxy approach. Known more formally as ‘methodological triangulation’, this approach is not used to validate results but, rather, is a strategy used in multiple academic disciplines to add “(…) rigor, breadth, complexity, richness, and depth to any inquiry” (Denzin 2012 p.81) and has shown to produce more robust and complete results (Thurmond 2001). However, this approach does not eliminate the biases inherent

14 to each method, which also need to be taken into consideration when interpreting results.

One of the strategies I applied throughout the data collection process was to check for intra-observer bias by repeating observations during different phases of the data collection, i.e. on different research trips separated by about a year to several months.

Methods can either be used together after the fact as Reed does with Makapansgat

(Reed 2013) or, preferably, at the time of data collection, as many recent paleoecological studies have done (e.g. Curran and Haile-Selassie 2016; Jones and Desantis 2017; Louys et al. 2012; Merceron et al. 2012; Sewell et al. 2018). The multi-proxy approach is particularly important for hominin paleoecology studies since some of the most commonly used proxies (e.g. carbon isotopes, mesowear, microwear) are indicators of paleodiet and thus cannot be used solely to reconstruct the realized niche of , including hominins. Other factors known to influence habitat selectivity in primates include predation risk, disease, weather, the presence of appropriate shelter or sleeping sites, access to water and activity patterns as well as various topographic and vegetative factors (distance between trees, size of trees, etc.) (Barton et al. 1992; Rovero and

Struhsaker 2007; Wong et al. 2006). Many of these factors can be studied through specific proxies.

1.5 Dissertation Objectives

This dissertation is divided in three chapters, in addition to the introduction (this chapter) and the conclusion (chapter 5):

15 Chapter 2 focuses on the site of Kanapoi. This site is the best studied of the three locations analyzed for this dissertation and is key to our understanding of the environmental context of Australopithecus anamensis. Various approaches have been used to reconstruct the environments at this site, and here I contribute new data and analyses using mesowear and hypsodonty. I analyzed the dental traits of 98 bovids, suids and rhinocerotids from Kanapoi. I evaluated mesowear and hypsodonty for each selected specimen. Results suggest that most of the animals analyzed had a relatively abrasive diet. Bovids in the assemblage also incorporated more grass into their diet than their modern counterparts. Results are also put into broader context and compared to all

Mio-Pleistocene African values known from the literature. Pliocene Kanapoi was most likely dominated by grassy environments, a conclusion that supports the results of previous investigations of the paleoecology of the site.

In Chapter 3, I focus on the most abundant taxon at most Pliocene East African fossil localities: Bovidae. I start by providing a thorough description of all bovid remains in the Allia Bay fossil assemblage, then I use a multi-proxy approach to describe and compare the paleoecology at Allia Bay, Kanapoi and Mursi by studying bovid specimens from each fossil assemblage using community ecology and taxonomy, carbon and oxygen stable isotopic values from dental enamel, mesowear and postcranial ecomorphology.

Analyses indicate that the bovid community composition is different at all three locations. Carbon isotopic values are generally more depleted at Mursi, corresponding to ingestion of C3 plants, which indicate closed settings. Intermediate enamel carbon isotopic values at Allia Bay and less depleted values at Kanapoi indicate the consumption of C4 plants, which suggests more open settings. Allia Bay mesowear scores are also

16 indicative of environments that are less open than at Kanapoi: bovids overall seemed to incorporate a more significant amount of browse into their diet. The ecomorphological analysis of the astragali also suggests a mixed habitat. The results correspond to a broad north-south environmental gradient, from a more closed habitat at Mursi to a more open environment in Kanapoi with intermediate environments at Allia Bay.

Chapter 4 further explores the paleoecology of all three fossil collections, but focuses on Suidae. I lay out the paleoecological implications for a faunal distribution that includes more suids than bovids, including the tendency for suids to be associated with more humid environments. I analyzed the data using faunal abundance patterns, carbon and oxygen stable isotopic values of the tooth enamel and surface modification patterns

(taphonomy). Suidae taxonomic abundance patterns are similar in the Kanapoi and Allia

Bay assemblages. Grazers such as Notochoerus are proportionally rarer in the Mursi assemblage than in the two other assemblages. In addition, I use carbon and oxygen stable isotopic data to compare dietary ecology in the suids from each assemblage. I use bone surface modification (taphonomy) observations to supplement the paleohabitat reconstruction of these taxa. Results from this chapter indicate that the paleoenvironments at Mursi are associated with humid settings. There, suids would have been able to thrive on varied diets and be able to stay in proximity to water. In addition, our results reinforce the ecological similarities between Allia Bay and Kanapoi, the only two sites where Australopithecus anamensis was found in the Omo-Turkana Basin.

In the final chapter, I tie together the conclusions from each chapter to provide a more complete reconstruction of the paleohabitats of what is arguably one of the most scientifically relevant hominin species. The results from the analyses undertaken in this

17 dissertation will impact the field’s understanding of the emergence and maintenance of obligate bipedal locomotion, a feature unique to the clade formed by humans and their closest relatives among primates.

Please note: All chapters are written as manuscripts, and I have indicated co-authors and contributions for each.

18

Chapter 2: Paleoecological implications of dental mesowear and hypsodonty in fossil ungulates from Kanapoi

Formatted in accordance with the guidelines of the peer-reviewed journal Journal of

Human Evolution, in review as of October 2016

Co-author: René Bobe

Co-author contributions: LD conceived of this study. LD drafted the manuscript with feedback from RB throughout the process. Both authors gave final approval for submission of this manuscript.

Keywords: mesowear; ungulate; paleoecology; Eastern Africa; Pliocene

19

Abstract

The Pliocene site of Kanapoi is key to our understanding of the environmental context of the earliest species of Australopithecus. Various approaches have been used to reconstruct the environments of this site, and here we contribute new data and analyses using mesowear and hypsodonty. The dental traits of ninety-eight bovids, suids and rhinocerotids from Kanapoi were analyzed using these proxies. Results indicate that most of the animals analyzed had a relatively abrasive diet. Bovids in the assemblage incorporated more grass into their diet than modern species of the same tribe or genus.

Although Pliocene Kanapoi likely had complex environments, our analysis indicates that grassy habitats were a dominant component of the ecosystem, a conclusion that supports the results of previous investigations of the paleoecology of the site.

Introduction

Understanding the environments of Kanapoi during the Pliocene is critical to understanding the ecology and adaptations of Australopithecus anamensis, the earliest species of the genus. Previous approaches include analyses of sedimentology (Feibel

2003), paleosols (Cerling et al. 2011; Wynn 2000), fossil vertebrate taxonomic composition (Bobe 2011; Geraads et al. 2013; Harris et al. 2003; Manthi 2008; Stewart

2003; Werdelin and Manthi 2012; Winkler 1998), and mammalian dental enamel stable isotopes (Cerling et al. 2015; Manthi et al. 2018). Although authors have reached somewhat different conclusions regarding Kanapoi paleoenvironments, the consensus seems to be that the site was dominated by relatively open, low tree-shrub vegetation

20 during the Pliocene. Here we contribute new data and analysis to test previous hypotheses of Kanapoi paleoenvironments.

Dental mesowear denotes the macroscopic dental wear facets created on the molars of ungulates during the lifetime of an . They vary in size and shape according to the properties of the foods ingested (abrasion) and the tooth-on-tooth wear during mastication (attrition) (Fortelius and Solounias 2000). Studies indicate that the shapes of the facets differ between grazing (grass-eating), browsing (leaf-eating) and mixed-feeding taxa. This is because of differences in the properties of the food types and the relative amount of grit they contain (Ungar 2015). Thus mesowear facets can be used to reconstruct the diets and therefore the environments in which these animals lived.

The relative height of the tooth crown, or hypsodonty, has often been used in conjunction with the evaluation of mesowear to reconstruct past habitats (Andrews and

Hixson 2014; Jernvall and Fortelius 2002) and it has been shown that using these methods in combination improves their accuracy (Fraser and Theodor 2011). Studies have shown that a large prevalence of high-crowned teeth in herbivores can be related to decreased rainfall (Eronen et al. 2010). This is because in a drier environment, the vegetation tends to be rougher and more abrasive. An animal’s teeth wear out as it feeds, and food items that contain grit and other abrasives wear down the tooth crown faster than softer foods, such as leaves. A tooth with a higher crown is therefore favorable in a drier environment because it will last longer. We will use the hypsodonty index to infer diet and environments in Kanapoi Pliocene mammals: a high hypsodonty index is likely to indicate a more grazing diet and feeding at ground level in open habitats, whereas a low one is linked to a more browsing diet (Damuth and Janis 2011). However,

21 hypsodonty is also a genetic trait linked to the “lifetime” of the tooth (Jernvall and

Fortelius 2002) which informs long-term evolutionary adaptation, and several studies have shown that it is not always directly indicative of the diet of the living animal

(Feranec 2003).

Mesowear and hypsodonty have been used increasingly in recent years to study questions related to human evolution. Mesowear was mentioned as one of the lines of evidence used in a multi-proxy framework to reconstruct the paleoenvironments at the As

Duma fossiliferous deposits of Gona (4.51-4.32 million years ago (Ma)), but results were not explained in details in the article (Semaw et al. 2005). At the upper Busidima

Formation (1.7 Ma - <0.64 Ma), Ethiopia, mesowear contributed to identifying two shifts in the paleoenvironment through time (Everett 2010). The approach was also used to demonstrate the presence of grass patches around 7 Ma at the site of Toros-Menalla in

Chad where Sahelanthropus tchadensis remains were found. The authors linked the presence of these grass patches to the development of bipedality in early hominins

(Blondel et al. 2010). White and colleagues (2009a) included mesowear in their multi- proxy reconstruction of the paleoecology of Ardipithecus ramidus at Aramis around 4.4

Ma. The resulting mesowear scores showed that browsing or frugivorous species dominated at the site. Similarly, Curran and Haile-Selassie used mesowear in their paleoecological analysis of Woranso-Mille (3.8-3.6 Ma), where remains of various

Australopithecus species were uncovered (Curran and Haile-Selassie 2016). The results from their mesowear analyses helped highlight the mosaic nature of the paleoenvironments at the site. Both hypsodonty and mesowear have been used to shed light on the paleoecology of Australopithecus africanus around 3.2-2.5 Ma at

22 Makapansgat, in South Africa (Schubert 2007). The authors ruled the hypsodonty results as unable to accurately reconstruct the diet in their study, but mesowear results correlated

well with stable isotope results and were further used to distinguish C3 browsers from C3 grazers at the site. Uno and colleagues (2011) report the results of a mesowear analysis on equids from the Nakali Formation and the Namurungule Formation at Samburu Hills during the Miocene, which correlate with the results of their stable isotope analyses. They documented a transition from browsing to grazing at the site starting at 9.9 Ma.

Mesowear studies have also been used more specifically on bovids at African sites from the Miocene to the Holocene. At Laetoli, mesowear was used to investigate dietary shifts and niche diversity in ungulates between two time periods separated by a million years: the Upper Laetolil Beds (ULB) at 3.85-3.63 Ma and the Upper Ndolanya

Beds (UNB) at 2.66 Ma (Kaiser 2011). Similarly, Faith and collaborators (2011) have also used this proxy to investigate the paleoecology of a previously poorly known bovid species from Rusinga Island in Kenya. Most recently, the identification of bovid remains from the Kibish Formation (~196-8 Ka), Ethiopia, included the evaluation of mesowear

(Rowan et al. 2015). The authors used this approach to document changes in dietary ecology in different bovid species from the Middle Pleistocene to the Middle Holocene, and demonstrated that dental abrasion levels were almost systematically higher in the past. Similarly, bovid mesowear data were used in conjunction with postcranial ecomorphology to reconstruct the paleoenvironments at the Omo Shungura Formation

(Barr 2014). The author thus documented multiple ecological shifts between geological members, which he argues is linked to either migration or eurytopicity in the Omo

Shungura hominins.

23

Finally, a recent large-scale ecometric analysis of the Turkana Basin over time showed that the west side of Lake Turkana as a whole was more arid than the east side starting around 4 Ma (Fortelius et al. 2016), about the age of the sediments of Kanapoi (Feibel,

2003). The present study provides a more localized ecological analysis focused specifically on the site of Kanapoi. The purpose of this article is to explore the paleoecology of Pliocene Kanapoi in the Turkana Basin using a ‘taxon-free’ approach to reconstruct habitat and diet of the large mammalian herbivores and to complement the existing data on the paleoecology of the site, particularly the results of Fortelius and colleagues’ recent investigation.

Material and Methods

Ninety-eight large herbivore specimens from the Kanapoi collection (1994-2015) were used for this study. All individuals belonging to the orders Artiodactyla, Perissodactyla and Proboscidea (APP taxa) (Eronen et al., 2010) were examined (Table 1 and

Supplementary Online Material). We only used teeth that were taxonomically identified to the genus level, or the tribe level for bovids. Only complete or nearly complete adult upper and lower molars were used. If more than one tooth of the same position was available for an individual, the average was used. Molars from older individuals (heavily worn) and subadults were not included. We used both isolated teeth and teeth still in place in the jaw. Unfortunately, no giraffids, equids or proboscideans met the criteria we set. The data on extant East African species is from Kaiser and colleagues (Kaiser et al.

2013). The bovid species in the modern dataset were gathered by tribe in order to make

24 the data more directly comparable to the fossil sample. We used published data from other sites relevant to hominin evolution in Africa to provide further contextualization of our results: the Kibish Formation (Rowan et al., 2015), Makapansgat (Schubert, 2007),

Laetoli (Kaiser, 2011), Toros-Menalla (Blondel et al., 2010) and Aramis (White et al.,

2009). We only selected data generated from second molars. When necessary, we also reframed the data using Kaiser et al. (2009)’s “0 to 4” summary system described below.

Taxon n %high %low %sharp %round %blunt HYP Category

Aepycerotini 14 7.14 92.86 0.0 50.0 50.0 2.41 m

Alcelaphini 11 63.64 36.36 45.45 45.45 9.09 2.25 m

Antilopini 8 12.5 87.5 50.0 25.0 25.0 1.94 m

Bovini 6 80.0 20.0 40.0 60.0 0.0 1.75 m

Hippotragini 4 25.0 75.0 25.0 75.0 0.0 1.59 m

Neotragini 7 42.86 57.14 71.43 14.29 14.29 1.51 m

Tragelaphini 15 53.33 46.67 53.33 46.67 0 1.41 b

Notochoerus 7 - - - - - 0.95 b

Nyanzachoerus 16 - - - - - 0.84 b

Rhinocerotidae 1 0 100 0 0 100 - -

Table 1: Mesowear and hypsodonty in Kanapoi ungulates. %high = percentage of teeth with high cusps, %low = percentage of teeth with low cusps, %sharp = percentage of teeth with sharp cusps, %round = percentage of teeth with rounded cusps, %blunt = percentage of teeth with blunt cusps, HYP = hypsodonty index, Category = hypsodonty index categories: b = brachydont, m = mesodont, h = hypsodont.

25 Mesowear

Mesowear (MSW) is scored according to two main criteria: cusp shape (sharp, rounded or blunt) and occlusal relief (high or low cusps) (Figure 3., Fortelius and Solounias,

2000). Following the original method, upper second molars (and ?M2) were used.

Similarly to Blondel and colleagues (2010), we also expanded our sample size to include lower second molars (and ?m2) (total n=68). It has been suggested that mandibular molars reflect a more grazing signal compared to the upper molars (Franz-Odendaal and

Kaiser 2003). In our sample, no significant differences between upper and lower molars were found at the p<0.05 level using a two-way ANOVA.

We are using the combined score as developed by Kaiser and colleagues (2009) in which the two traits are converted into mesowear stages from 0 to 4. This combined score is also used to describe the modern comparative dataset used here (Kaiser et al. 2013).

The score can be interpreted as the combined effect of abrasion and attrition on the teeth.

Mesowear scores are useful for comparative purposes, but their downside is that they muddle the relative contributions of both phenomena. Animals that mostly feed on leaves will exhibit higher and sharper cusps than those that feed on grass. In browsers, the wear is caused mainly by tooth-on-tooth contact. A stage “0” (sharp and high cusps) is a hyper browser. Grazers tend to have blunt or round cusps that tend to be low. However, some grazers also exhibit high relief (Fortelius and Solounias 2000). The shape of their teeth is influenced mostly by the action of tooth-to-food wear during mastication. We interpret a stage “4” (blunt cusps and low relief) as hyper grazer. Mesowear has been tested for inter-observer and intra-observer biases, which were found to be negligible (Kaiser et al.

2000).

26

Figure 3: Illustration of traits used in the evaluation of mesowear and hypsodonty. (A) Tragelaphini, KNM-KP 58605, mesowear score (MSW) 0: sharp and high; (B) Tragelaphini, KNM-KP 30421, MSW 1: round and high; (C) Tragelaphini, KNM- KP 58607, MSW 2: sharp and low; (D) Tragelaphini, KNM-KP 56847, MSW 3: round and low; (E) Alcelaphini, KNM-KP 58608, MSW 4: blunt and low. All specimens are from the Kenya National Museums, Nairobi.

Hypsodonty

Hypsodonty (HYP) was evaluated using the ratio of height to width for every lower third molar in the collection (Figure 3, Janis, 1988). Then, lightly worn lower third molar (m3s and ?m3s) from the Kanapoi assemblage were classified by tribe or genus (n=50).

Some teeth are from the same individuals as the teeth used for the mesowear analysis (n=19), but most are not. A note about these particular individuals is included in the Results section. Next, the tribe or genus average was computed and classified as one of three possible categories: individuals are categorized as mesodont when their hypsodonty ratio is between 1.5 and 3, brachydont when the score is less than 1.5, and hypsodont when it is higher than 3 (Damuth and Janis 2011).

Results

27 The teeth of bovids from the Kanapoi collection, dated to ~4 Ma, were clearly affected by a diet containing abrasive elements, as seen in the relatively high mesowear scores in

Figure 4. There was a statistically significant difference between tribes across all sites according to an ANOVA, F(6, 41)= 4.018, p=0.00296. Pairwise comparisons of the means according to Tukey HSD’s test showed significant differences between three pairs:

Tragelaphini-Hippotragini (p=0.01), Neotragini-Hippotragini (p=0.02) and Tragelaphini-

Alcelaphini (p=0.04). There were also significant differences between sites (ANOVA,

F(9, 38)= 2.487, p=0.024), but the differences are not maintained when examined through pairwise comparisons at the p=0.05 level using Tukey HSD’s test. The ANOVA is not significant at the p=0.05 level when the site or the period is used as an interactive factor in addition to tribal differences.

There is a moderate correlation between hypsodonty and mesowear in bovids in our data set (Pearson: r = 0.57). This is expected since both a higher hypsodonty index and a higher mesowear score are indicative of a grazing diet. We suggest that the lack of a strong correlation between hypsodonty and mesowear in our bovid sample can be explained by the difference in scale between the two proxies. Hypsodonty is strongly influenced by genetics and does not necessarily reflect the current diet of an animal in a given environment (Davis and Pineda Munoz 2016). Nineteen specimens consisted of individuals with at least a second molar suited for the mesowear analysis and a third molar that could be used to calculate hypsodonty. There is a weak positive correlation between hypsodonty index and mesowear score when measured in these specimens

(Pearson: r= 0.28, Figure 5). In addition to also being biased by the different scales of the

28

Figure 4: Distribution of mesowear scores in the Kanapoi fossils, extant eastern African bovids (Kaiser et al. 2013) and other fossil localities: Aramis (White et al., 2009), Kibish (Rowan et al., 2015), Laetoli (Kaiser, 2011), Makapansgat (Schubert, 2007) and Toros-Menalla (Blondel et al., 2010).

30

Figure 5: Distribution of the mesowear score and hypsodonty index for the 19 Kanapoi bovid specimens for which both measurements are possible.

proxies, the fact that this correlation is less strong than for the tribal averages reinforces the notion that proxies like mesowear and hypsodonty need to be used at the assemblage level, which reduces the influence of individual variation. With such a low sample size, it is difficult to discern any tribal pattern, but this is an interesting area of future investigation if the sample size was to increase.

Fossil Aepycerotini had an attrition-dominated diet at Kanapoi. With an average of 3.36 ± 0.84, they show the strongest grazing signal in the assemblage. In addition, aepycerotins have the highest hypsodonty indices among the Kanapoi fossils

(HYP=2.41), which is expected in grazing animals. They also have the lowest standard deviation in the sample, indicating low within-tribe variability. This is not surprising, considering that all individuals scored have been attributed at least tentatively to a single genus (Aepyceros sp. (n=12). or cf. Aepyceros (n=2)) (Geraads and Bobe 2018a). The

31

enamel of three specimens used in this study, KNM-KP 32546, 29259 and 32823, have also been analyzed isotopically and show grazing or graze-dominated mixed diets

(Cerling et al. 2015). Amongst the fossil Aepycerotini included here, Kanapoi aepycerotins stand out as having a much more grazing signal, completely out of the range of the individuals from the same tribe at other sites. Other fossil collections show much more browsing signals for the aepycerotin second molars, the highest being at Kibish member I (MSW=1.33) and the lowest being the Upper Laetoli Beds (MSW=0.24)

(Kaiser 2011; Rowan et al. 2015) (Figure 4). As reflected in their mesowear score

(MSW=0.98), even modern impalas (Aepyceros melampus) are classified as abrasion- dominated mixed feeders (Cerling et al. 2015; Franz-Odendaal and Kaiser 2003).

Fossil antilopins at Kanapoi are mixed-feeders that incorporated some graze into their diets and have an average mesowear score of 2.5 ± 1.31. They tend to have low cusps that are sharp or rounded. They also have relatively hypsodont teeth (HYP=1.94).

Among the antilopins analyzed, specimens were identified as Gazella sp. (n=4), cf.

Gazella (n=1) or indeterminate (n=3). Gazella cf. janenshi and a species with affinities to

Dytikodorcas sp. are additional antilopins identified within the Kanapoi fossil assemblage but could not be analyzed for this study (Geraads and Bobe 2018a). Antilopin mesowear results overlap with those of their tribe members inhabiting the earlier site of Toros-

Menalla, in Chad (Blondel et al. 2010). On the contrary, the tribe has an attrition- dominated diet at Laetoli (MSW= 0.37-0.62) and Makapansgat (MSW=0.52) (Kaiser

2011). In addition, the modern members of this tribe, the gerenuk (MSW=0.87), Grant’s gazelle (MSW=0.65) and dibatag (MSW=0.71), all have much more of a browsing signals despite being described as browse-dominated mixed feeders in the behavioral

32

literature (Gagnon and Chew 2000). This is also confirmed by isotopic studies of animals in the wild (Cerling et al. 2015; Ngugi et al. 2014).

Kanapoi Hippotragini have rounded or sharp cusps and the cusps tend to be low, which is interpreted as a mixed-feeding signal. Their hypsodonty score is 1.59 and their mesowear score is 2.25 ± 1.5. They also have the highest standard deviation in the data set, denoting variability within the tribe, a characteristic also detected by isotopic studies at Laetoli (Kingston and Harrison 2007). However, hippotragins are rare in the assemblage and it is unclear how many species there were at Kanapoi. All remains are attributed to either cf. Tchadotragus (n=1) or indeterminate (n=3) (Geraads and Bobe

2018a; Geraads et al. 2013). Amongst fossils, hippotragins from the Chadian site of

Toros-Manella have the mean mesowear score the most similar to that of Kanapoi

(MSW=1.34), but their signal is more attrition dominated (Blondel et al. 2010).

Hippotragins have a browsing diet at the Upper Laetoli Beds (MSW=0.6) (Kaiser 2011), but a much more graze-dominated diet throughout all members of the Kibish formation

(MSW=3.5 to 4) (Rowan et al. 2015), a testament of the tribe’s dietary flexibility. The modern Hippotragini used in Kaiser and colleague’s study (Kaiser et al. 2013), the roan

(Hippotragus equinus) and sable antelopes (Hippotragus niger), are classified as mixed feeders by the mesowear analysis. The individuals sampled incorporated some browse in their diets, with an average mesowear score of 1.44 ± 0.21. This is particularly unexpected for the roan antelope, typically a more grazing taxon (Gagnon and Chew

2000). Studies have shown that this species’ diet tends to vary substantially throughout the year (Havemann et al. 2016). It is thus possible that the relatively large difference between the fossil and living specimens is attributable to the fallback foods integrated in

33

the diet of the individuals sampled. Their teeth are also relatively hypsodont (Kaiser et al., 2013), which is more in line with what is known about the behavioral record of the members of this tribe.

Neotragins form a paraphyletic clade (Bibi et al. 2009; Matthee and Robinson

1999), but we use this nomenclature here to refer to ecologically similar small sized antelopes that cannot be classified as Antilopini, as is typical in paleoecological literature.

The specimens used here are classified as either Raphicerus (n=1), cf. Raphicerus (n=2) or indeterminate (n=4) and no additional neotragin taxon has been identified in the

Kanapoi collection in general (Geraads et al. 2013). Their hypsodonty index is 1.51, one of the lowest amongst Kanapoi bovids. With an average mesowear score of 1.57 ± 1.4, they are classified as attrition-influenced mixed feeders. This score was the most similar to that of neotragins from the Member I of the Kibish Formation (MSW= 1.22) (Rowan et al. 2015). Fossil Neotragini from the Laetoli, Woranso-Mille, the Kibish Formation

Member III and Aramis as well as the modern species (steenboks and klipspringers) are obligate browsers (MSW< 1), as confirmed by observational studies in the wild (Curran and Haile-Selassie 2016; Kaiser 2011; Kingdon et al. 2013b; Rowan et al. 2015; White et al. 2009a). A similar conclusion was reached about the Laetoli neotragins in a study of their isotopic signals (Kingston and Harrison 2007). In sum, the Kanapoi neotragins would have grazed more frequently than any other neotragin included in this study, including the modern species.

Similarly, the results indicate that fossil tragelaphins display a mixed, but attrition-influenced signal, with an average of 1.4 ± 1.12. Their hypsodonty index is 1.41 and is the lowest in the assemblage. Tragelaphini are the only bovids in the Kanapoi

34

assemblage classified as brachydont. None of the specimens used here could be identified beyond the tribe level, although Tragelaphus cf. kyaloi is currently the only tragelaphin recognized at Kanapoi (Geraads and Bobe 2018a). Isotopic studies have shown the presence of a mixed, but attrition dominated diet in tragelaphins during the Plio-

Pleistocene (Cerling et al. 2015). This represents a major difference from their modern counterparts. The greater and lesser kudu and the bushbucks are browsers (Cerling et al.

2015; Gagnon and Chew 2000). Elands typically have a more flexible diet, switching from grazing to browsing between seasons (Gagnon and Chew 2000). But, together, the modern Tragelaphini have a browsing mesowear signal (0.6 ± 0.27). Tragelaphins are also considered browsers at most of the other fossil sites for which mesowear data is published. The two individuals studied in the Laetoli collection are a hyper browser

(MSW=0) in the Upper Laetoli Bed and an attrition influenced mixed feeder (MSW=1.2) in the Upper Ndolanya Beds (Kaiser 2011). The tragelaphins from the earliest member of the Kibish Formation have an average mesowear score of 0.25, which also corresponds to a browsing signal (Rowan et al. 2015). Tragelaphins are also considered browsers at

Aramis (MSW=0.5) (White et al. 2009a), Woranso-Mille (Curran and Haile-Selassie

2016) and Makapansgat (MSW=0.24) (Schubert 2007).

With an average of 1.3 ± 1.4, the alcelaphin fossils from the Kanapoi collection are interpreted as having a mixed diet, which incorporates some browse. This is similar to the values seen in alcelaphins from the Laetoli succession (MSW at ULB=0.82 and

UNB=1.84), Aramis (MSW=1) and Woranso-Mille (Curran and Haile-Selassie 2016;

Kaiser 2011; White et al. 2009a). The values for the Kibish Formation (MSW= 3 to 3.33) and Makapansgat (MSW= 3.4) stand out as displaying a definite grazing signal in

35

members of this tribe (Rowan et al., 2015; Schubert, 2007). At the same time, alcelaphins have one of the highest hypsodonty levels within the fossil bovid sample, a trait related to a grazing diet (HYP=2.25). None of the Alcelaphini specimens (n=11) we analyzed have been identified beyond the tribe level, despite the fact that at least four distinct species have been recognized in the Kanapoi collection: Damalacra harrisi, cf. Damalborea n. sp. and two additional ones which have not been formally attributed to a taxon (Geraads and Bobe 2018a). Alcelaphini is also one of the rare tribes that show increased grazing in more modern lineages. The same relationship is also shown through isotopic analyses

(Cerling et al. 2015). Although the average mesowear signal of the modern Alcelaphini species is that of a mixed-feeder (MSW= 2.07), the black wildebeest, the blue wildebeest, the and the tsessebe tend to be described in the ecological literature as grazers or variable grazers (Gagnon and Chew 2000; Kingdon et al. 2013b).

And finally, fossil Bovini display a browsing signal, with a score of 1±1.1, which is very close to the signal of 1.17 of the extant African buffalo (Syncerus caffer) (Kaiser et al., 2013). All the teeth attributed to Bovini have been identified to either Simatherium sp. (n=3) or indeterminate (n=4), as is the case for all Bovini remains recovered so far at

Kanapoi (Geraads and Bobe 2018a). The mesowear analysis scores can seem unexpected both for the extant and extinct specimens of this tribe. However, in their description of the original mesowear method, Fortelius and Solounias (2000) note that cusp sharpness is the variable that specifically reflects abrasion. Mesowear studies have consistently revealed high and rounded cusps for both extant and extinct African Bovini (Blondel et al. 2010; Curran and Haile-Selassie 2016; Fortelius and Solounias 2000; Kaiser 2011;

Rivals et al. 2007; Rowan et al. 2015), which is also the case for Kanapoi Bovini fossils.

36

The scores obtained illustrate the downside of using the mesowear scoring method, which muddles the effects of abrasion and attrition. A similar pattern is observable in another

Bovini, the American bison (Bison bison) (Rivals et al., 2007). The authors of the study acknowledge the possibility that the American fossil Bovini were integrating browse into their diet. However, Rivals and colleagues further suggest that the low mesowear signature might also be explained in part by the consumption of different types of grass or to a particularly low level of abrasive particles in the food the animals select. A similar conclusion was reached for Bovini at Laetoli (Kaiser 2011). In summary, Kanapoi Bovini were likely grazers and other proxies support this. Isotope analyses have classified both fossil and extant Bovini as grazers (Cerling et al., 2015). Similarly, the fossil Bovini from

Kanapoi have a moderate hypsodonty index (HYP=1.75). Furthermore, the living African buffalo is typically classified as a “variable grazer”, although studies have shown that their diet can differ significantly with geography and seasonality (Gagnon and Chew

2000; Kingdon et al. 2013b).

Suids at Kanapoi are all brachydont. Notochoerus jaegeri has more hypsodont teeth (HYP=0.95) than Nyanzachoerus kanamensis (HYP=0.84), which is indicative of a more grazing diet. Isotopic analyses reveal that both fossil genera were mixed-feeders that included graze into their diets (Cerling et al. 2015; Harris and Cerling 2002). A recent study of the microwear of the molars of both genera present similar results (Ungar et al. 2018). Modern East African suids (Potamochoerus, Hylochoerus) are mostly versatile mixed feeders that integrate browse into their diets, with the exception of the grazing Phacochoerus (Kingdon et al. 2013b). There is currently no suid-focused

37

mesowear method. The development of such a method could add nuance to these results but is out of the scope of this paper.

Only one Rhinocerotidae could be studied for this paper. The tooth belongs to a

Ceratotherium (white rhinoceros) specimen. The rhinoceros has blunt and low cusps, and thus is considered an extreme grazer with its score of 4. This sample is too small to attribute patterns to the genus. The modern Ceratotherium is indeed a grazer (Kingdon et al. 2013a).

Discussion

The mesowear and hypsodonty data presented here indicate that the environments at

Kanapoi were relatively grassy and dominated by ungulates with a grazing or mixed diet.

This conclusion is similar to that of the study of the dental microwear in Kanapoi bovids

(Ungar et al. 2018). Notably, the sites closest in age and geography to Kanapoi and for which mesowear data are published – Aramis, the Upper Laetolil Beds and Woranso-

Mille – show generally a more browsing diet for each bovid tribe. The bovids from

Kanapoi had frequent overlaps with those of Toros-Menalla, a site with relatively open paleoenvironments including gallery forests, open grasslands and even desert conditions

(Vignaud et al. 2002).

Aepycerotins were the most grazing taxon at Kanapoi, but were either browse- dominated intermediate feeders or browsers at every other site included in this study.

Antilopins are systematically categorized as browsers in the African Plio-Pleistocene record, and only incorporate graze into their diets at Kanapoi, and possibly at Woranso-

Mille. In both hippotragins and alcelaphins, Kanapoi bovids are at the center of the

38

distribution. Kanapoi tragelaphins had similar diets to those of Aramis and Laetoli

(UNB). Both neotragins and tragelaphins were classified as browsers at most sites except

Kanapoi, where they are characterized as browse-dominated mixed-feeders. Bovini in this study are systematically classified as mixed-feeders or browsers (although, as we discussed, the signal may lead to different interpretations).

However, the mesowear results of the fossil specimens in our sample all have large standard deviations. This denotes significant within-tribe variation, which may be due to variation between the different genera or species within the tribe, as well as to individual variation. In addition, there is some discordance between the diet of living taxa as described in the behavioral literature and their mesowear score, which we attribute to the underestimation of the importance of fallback foods in the diets of these animals in the behavioral literature. In addition, at least in the case of Bovini, the mesowear summary scoring system showed limitations: it muddled the relative contributions of attrition and abrasion to the mesowear score. The two suid genera, Notochoerus and

Nyanzachoerus, seem to integrate grazing in their diet.

Furthermore, bovids at Kanapoi are clearly dominated by taxa with a more abrasion- dominated diet than their modern counterparts. These results echo the study by Rowan and colleagues (2015) of the Kibish bovids, which suggests that modern taxa recently began integrating more browse in their diets. Our results also support Fortelius and colleagues’ portrayal of West Turkana as having relatively open vegetation around 4 Ma

(Fortelius et al., 2016). The results of Cerling and colleagues’ stable isotope study of dietary changes in APP taxa in the Turkana Basin from 4 Ma to the present also closely mirrors our own findings (Cerling et al., 2015).

39

Conclusion

Our results corroborate previous paleoecological analyses of the Pliocene Kanapoi site using different methods. Previous studies have demonstrated the presence of Dite paleosols as well as open-adapted mammals at the site, and thus the paleoenvironments and vegetation at Kanapoi have been reconstructed as being heterogeneous, but relatively open (Geraads et al. 2013; Ungar et al. 2018; Wynn 2000). Renewed fieldwork would allow the accumulation of more fossil specimens, which would be particularly helpful to expand the sample size of Rhinocerotidae as well as include other taxa such as Giraffidae in the analysis of dental mesowear. One major caveat of this study is that the taxonomic resolution of the data did not allow us to analyze bovids beyond the tribe level, as is recommended in the literature (Louys et al. 2015). Additional analytical methods such as post-cranial ecomorphology would also help refine the results of this study.

Nevertheless, the results presented here contribute new data to our understanding of early Australopithecus environments. Although Pliocene environments at Kanapoi were undoubtedly complex, dental mesowear and hypsodonty in Kanapoi APP mammals indicate that the geographic distribution of early Australopithecus included fairly open and grassy environments. Despite this, the !13C values of the enamel of A. anamensis indicate that this species had a diet dominated by C3 plants, whereas later hominins

exploited a much larger range of resources, which included both C3 and C4 plants

(Sponheimer et al. 2013). C3 resources were undoubtedly available at Kanapoi. Dental characteristics of A. anamensis such as their post-canine megadontia, thick enamel, robust mandibles and dental microwear signature, demonstrate that these early

australopiths likely exploited harder and more abrasive food within the C3 spectrum

40

(Estebaranz et al. 2012; Macho et al. 2005; Ungar et al. 2018; Ward et al. 2001). The realized niche of all animals, including hominins, cannot solely be explained by dietary preferences. Other factors known to influence habitat selectivity in primates include predation risk, disease, weather, the presence of appropriate shelter or sleeping sites, access to water and activity patterns as well as various topographic and vegetative factors

(distance between trees, size of trees, etc.) (Barton et al. 1992; Rovero and Struhsaker

2007; Wong et al. 2006). We can thus suggest that A. anamensis perhaps inhabited

Kanapoi for reasons linked to other aspects of their ecology, such as locomotion. Further research on paleoenvironments at additional A. anamensis-bearing sites such as Allia Bay

(Koobi Fora Formation, Kenya) is necessary to shed light on the breadth of environments inhabited by hominins around 4 million years ago.

Acknowledgements

We are grateful to the two anonymous reviewers for their helpful suggestions. We thank the National Museum of Kenya staff for facilitating access to the specimens. Merci D.

Geraads for the updated taxonomic identification of the specimens. Finally, thank you to

F.K. Manthi, C.V. Ward and J.M. Plavcan who lead the West Turkana Paleo Project.

Funding: This research was funded by a dissertation research grant from the Leakey

Foundation to LD, Sigma Xi Grants-in-Aid of Research, Explorers Club Washington

Group inc, Evolving Earth Foundation, Cosmos Club Foundation and the Lewis N.

Cotlow Fund. Funding for the Kanapoi fieldwork is from NSF BCS-1231749 (to CVW &

FKM), NSF BCS-1231675 (to JMP& PSU), the Wenner Gren Foundation as well as the

University of Missouri Research Board.

41

Chapter 3: Multi-proxy analysis of fossil Bovidae (Mammalia,

Cetartiodactyla) from the Middle Pliocene Allia Bay, East Turkana, Kenya and implications for Australopithecus anamensis paleoecology

In preparation for submission in Summer 2018

Co-authors: René Bobe, Jonathan G. Wynn, W. Andrew Barr

Co-author contributions: LD conceived of this study. RB and WAB contributed to the design. LD drafted the manuscript with feedback from all authors throughout the process.

Keywords: bovid; paleoecology; Eastern Africa; Pliocene; stable isotopes

42

Multi-proxy analysis of fossil Bovidae (Mammalia, Cetartiodactyla) from the

Middle Pliocene Allia Bay, East Turkana, Kenya and implications for

Australopithecus anamensis paleoecology

Abstract

Australopithecus anamensis, possibly the earliest fully bipedal hominin species, lived in eastern Africa around 4 million years ago. Fossil remains associated with this species are relatively rare, and have only been found at a handful of sites. We present here the first multi-proxy paleoecological analysis from one of these sites: Allia Bay, Koobi Fora

Formation, East Turkana, Kenya (~3.98 Ma). This paper focuses on the family Bovidae, a taxon that is abundant and ecologically diverse during the Pliocene in eastern Africa. We use multiple lines of evidence to derive a multidimensional reconstruction of the Allia

Bay paleoenvironments represented by the fossils studied: taxonomy, postcranial ecomorphology, enamel oxygen and carbon isotope values as well as mesowear and hypsodonty index. We compare our results to the two localities of similar age in the

Omo-Turkana Basin: Kanapoi, Kenya, and Mursi, Ethiopia. The majority of the fossils attributed to A. anamensis have been found at Kanapoi (c.70%), some have been discovered at Allia Bay (c. 30%) (Ward et al. 2013) and none have been found so far at the Mursi Formation. Results show that the bovid community composition is significantly different between the three fossil assemblages. Astragalus ecomorphology suggests a mixed habitat at both Kanapoi and Allia Bay. δC13 values at Allia Bay compared between members of the same tribe are broadly similar to those from Kanapoi. δC13 values are

43

indicative of C3-browsing diets, reflecting closed environments at Mursi. Allia Bay dental mesowear scores are also similar to results from Kanapoi. Overall the evidence from dental enamel carbon and oxygen data, mesowear and hypsodonty of bovids indicate greater amount of browse into their diet at Mursi compared to Allia Bay and Kanapoi.

These data suggest more closed vegetation structure at Mursi and more open but heterogeneous environments at Allia Bay, which is similar to the environments reconstructed for Kanapoi. Australopithecus anamensis thus seems associated with open but heterogeneous settings rather than more closed environments such as those present at the Mursi locality, which is in line with their bipedal locomotion and other adaptations characterizing the hominin lineage.

Introduction

Australopithecus anamensis, which is the earliest Australopithecus species, lived in the

Omo-Turkana Basin c.4 million years ago (Ma). Compared to other hominin species, A. anamensis paleoecology is poorly known, and relies heavily on data from Kanapoi, the site from which most A. anamensis specimens are described (69 of 100 total specimens as of 2018). It is currently unknown if the environments that A. anamensis inhabited at

Kanapoi are representative of the habitats of the species in the Omo-Turkana Basin as a whole (Figure 1). Tibial remains (KNM-KP 29285) attributed to this species from

Kanapoi posses traits that have been interpreted as adaptations to committed bipedalism, including a knee directly in line with the foot (Leakey et al. 1998; Ward et al. 2001). The remaining 31 fossil remains attributed to A. anamensis recovered from the Omo-Turkana

44

Basin were found in locality 261-1 of the Koobi Fora Formation in Kenya. Detailed description and analyses of these hominin fossils were produced (Grine et al. 2006;

Leakey et al. 1998; Macho et al. 2005; Ward et al. 1999), in addition to primate remains from the same locality (Jablonski et al. 2008) and carbon and oxygen isotope values from a subset of the fauna (Blumenthal et al. 2017; Cerling et al. 2015; Schoeninger et al.

2003). However, the remaining ~2,000 vertebrate fossils from this hominin-bearing site recovered in the 1990s have yet to be formally described. These unpublished fossils have the potential to significantly impact our understanding of the beginnings of the

Australopithecus lineage.

About 800 fossils have been recovered from the Mursi Formation and 256 of them were described in a summary article (Drapeau et al. 2014), but no hominin remains were found. The fossils are estimated to be at least 4 million years old, since the overlying Mursi Basalt in dated to 4.2 Ma (Haileab et al. 2004). Suids and aquatic taxa are abundant at Mursi, and the paleoenvironments are reconstructed as relatively closed mosaic habitats. Yet, ecological reconstructions using other proxies (fossil wood, isotopes) suggest that more open environments were also present at this locality (Drapeau et al., 2014).

Most of what is known about the paleoecology of A. anamensis can be attributed to researchers analyzing fossils from the site of Kanapoi. Previous studies have shown the presence of Dite paleosols suggesting semi-arid environments as well as open-adapted mammals, and thus the paleoenvironment and vegetation at Kanapoi have been reconstructed as heterogeneous, but relatively open (Geraads et al. 2013; Harris et al.

2003; Ungar et al. 2018; Wynn 2000).

45

This study presents a multi-proxy study of the Bovidae from Allia Bay and compares them to Bovidae from Mursi and Kanapoi. Using multiple lines of evidence in combination is an approach increasingly used in paleoecology (Blondel et al. 2010;

Curran and Haile-Selassie 2016; Reed 2013; Sánchez-Hernández et al. 2016) as it allows for a multidimensional reconstruction of a site. In addition, this approach circumvents certain biases, such as the different time and geographic scales to which the results from each of these methods correspond

(Bobe et al. 2007a; Davis and Pineda Munoz 2016). We specifically focus on bovid remains because they are the most abundant taxonomic family in the Allia Bay assemblage, and due to their habitat-specific adaptations, they are competent proxies for reconstructing ancient vegetation, which is a key aspect of habitat preferences (Bibi et al.

2009; Bobe 1997; Bobe and Eck 2001; Vrba 1995).

Background

Context

Australopithecus anamensis fossils were recovered from the Lonyumun Member of the

Koobi Fora Formation in the Allia Bay region, locality 261-1 (3°38' N and 36°16' E), hereafter referred to as “Allia Bay” (Figure 2). Although Allia Bay was recognized as a fossil-bearing locality in early prospections of the Koobi Fora Formation, Feibel (1988) was the first to formally identify and name the locality. He describes the sediments as sandstones containing highly polished fossilized bones and teeth. The Allia Bay fossils

46

were deposited in a fluvial context and represent a single depositional event (Coffing et al. 1994). They were found in a bone bed situated below or within the Moiti Tuff, itself dated to 3.97+/- 0.03 (McDougall & Brown, 2008). The bone bed is about 200 m2 in extent and 20 cm thick and is estimated to be 3.98 Ma (Coffing et al. 1994; McDougall and Brown 2008).

A few fossil specimens are surface finds collected by the initial reconnaissance expedition of the Koobi Fora Formation in 1968 (Harris 1983a). Locality 261-1 was then explored for fossils during the 1980s and 1990s by the Koobi Fora Field School led by the National Museums of Kenya and Harvard University. Then, it was thoroughly excavated by a team from the National Museums of Kenya led by Dr. Meave G. Leakey between 1995 and 1997. Thirty one hominin fossils, most of which are cranio-dental specimens as well as numerous faunal remains were recovered from the excavations

(Ward et al. 1999).

Taphonomy

There are known differences between Allia Bay and Kanapoi in terms of taphonomy and differential depositional environments at the two sites. One of the main impacts of divergent depositional environments is bone transport. For instance, denser and more resistant bones are more common in high-energy fluvial depositional environments, because the lighter and less dense elements tend to be transported away from the site by water (Vorrhies 1969). Other factors to consider are space and time-averaging, especially relevant here because of the localities of Allia Bay and Mursi are more constrained in

47

time (Behrensmeyer and Reed 2013), and collection methods. Allia Bay fossils were collected from the surface in 1981 and 1987, and then the site was excavated between

1995 and 1997. Mursi was only explored through surface collections, but particular care was given to collect every specimen found, even unidentifiable fragments (Drapeau, pers. comm.). Early work at Mursi took place in the 1960s and 1970s and, more recently, between 2009 and 2014. Notably, fieldwork was limited at the Mursi Formation during the 1960s and 1970s because access to exposures was difficult (Drapeau et al. 2014). The

Allia Bay bonebed was completely excavated and a portion of the site was sieved using a

2 mm mesh. Over 2,000 faunal remains, the vast majority being unidentifiable non-mammals, were recovered during this process and described in a Master’s thesis

(Hagemann 2010). The much larger Kanapoi exposures have been explored thus far through excavation, bone walks and surface collections. Fossils were surface collected in the 1960s by a team led by Bryan Patterson (Patterson and Howells 1967). All three methods were used in the 1990s by a team led by Meave G. Leakey (Leakey et al. 1995;

Leakey et al. 1998) and again since 2003 by teams led by Fredrick K. Manthi, Carol V.

Ward and J. Michael Plavcan. Kanapoi is also particularly rich in hominin remains and is associated to many recent paleontological and paleoecological publications (Boisserie

2018; Brochu 2018; Delfino 2018; Field 2018; Geraads 2018; Geraads and Bobe 2018a;

Geraads and Bobe 2018b; Manthi et al. 2018; Ungar et al. 2018; Van Bocxlaer 2018;

Werdelin and Lewis 2018).

48

Materials

Here we focus on the 648 fossils assigned to Bovidae within the Allia Bay collection.

This includes 278 post-cranial remains, 50 horn cores and frontlets, 23 mandibles or maxillae with teeth in place as well as 281 isolated teeth. In this sample, 364 fragmentary specimens were identified only to the family level and are not described further. The remaining 284 specimens are identified to the tribe level. Specimens attributed to species and other noteworthy specimens are described in the next sections. The fossil specimens from Allia Bay and Kanapoi are housed at the Nairobi National Museum in Kenya. The

Mursi specimens are kept at the National Museum of Ethiopia in Addis Ababa, Ethiopia.

Over half of the Allia Bay specimens are complete (44.5%) and an additional

10.8% consists of “mostly complete” elements. We consider a specimen as “mostly complete” when at least 75% of the surface of the element can be observed (i.e., not missing or enclosed in the matrix). A jaw must possess both bone and teeth to be classified as complete. The remaining specimens are all fragments, which are defined as comprising <50% of the original element. The Mursi bovid collection consists mostly of fragmentary specimens. Within the Kanapoi collection, complete and incomplete elements (either “mostly complete” or fragments) are about equal in abundance. Within fragmentary remains, size distribution also differs between fossil collections from each site. Although the three assemblages are composed mainly of small size fragments (<5 cm), specimens from Allia Bay (mode = 2.0-2.9 cm) are generally larger than those from

Kanapoi (mode = 1.0-1.9 cm).

49

The nomenclature used here follows Grubb (Grubb 2001). The measurements are given in millimeters. Dental buccolingual (or “width”) and mesiodistal (or “length”) measurements are taken mid-crown. Horn core measurements are summarized in Table 2 and dental measurements are presented in Table 3.

Abbreviations

AB, Allia Bay; DT, transverse diameter of the basal horn core; DAP, anteroposterior diameter of the basal horn core; ER, East Rudolf (= East Turkana); KP, Kanapoi; KNM,

Kenya National Museum; Lt., left; LT, Lothagam; Rt., right; WT, West Turkana

Systematic Paleontology

CETARTIODACTYLA Montgelard, Carzeflis, and Douzery, 1997

BOVIDAE Gray, 1821

BOVINAE Gray, 1821

TRAGELAPHINI Blyth, 1863

TRAGELAPHUS De Blainville, 1816

A total of 132 remains are assigned to the tribe Tragelaphini (or cf. Tragelaphini).

Tragelaphin spiraled horn cores are relatively triangular in cross-section and possess an anterior keel (Gentry and Gentry 1978; Harris et al. 2003) (Table 2). Tragelaphin molars from Allia Bay are medium to large in size, with faint ribs and no goat fold. The upper teeth are particularly large, with rounded styles. Some specimens (e.g. KNM-ER 43397)

50

have pinched lingual or labial lobes. The central enamel islands have the shape of an open V. The M1 and M2 have constricted and mesiodistally elongated lobes, with the mesial lobe distinctively elongated compared to the distal lobe. Most teeth in the collection are very worn, which may be due to a highly abrasive diet or an abundance of older individuals. Some very worn M3 specimens have a metastyle that presents a significant distal projection and forms a 90-degree angle with the body of the tooth (e.g.,

KNM-ER 43394, Figure 6K). The M3 also have particularly large roots and some are also characterizes by bulbous lobes (e.g. KNM-ER 43384).

The paraconid and metaconid of the mandibular premolars are unfused. On these teeth, the paraconid is differentiated from the parastylid. In most specimens, the P4 hypoconid is hardly salient. The lower molars are relatively small sized with strong to moderate central pillars, which may or may not be connected to the occlusal surface of the tooth. They have small central enamel islands, which have the shape of an open V.

The lingual ribs are angular, but not particularly salient and the styles are weak. The labial lobes are generally set far apart from one another and triangular to subtriangular

and elongated. The M3 has a salient hypoconulid that projects distally. There is a large amount of variation, which we attribute to the coexistence of multiple species at Allia

Bay c.4 Ma.

KNM-ER 42687 and 42842 are both hemi-mandibles with relatively complete tooth rows, which illustrate the amount of variation within the tribe. KNM-ER 42687 is a

right hemi-mandible with an almost complete tooth row, from P4 to M3 (Figure 6J). The

M1 is particularly small and possesses a basal pillar connected to the occlusal surface of

51

the tooth. The other molars also have weak basal pillars, but they are disconnected from the occlusal surface. The M3 has a relatively strong and posteriorly projecting hypoconulid. The molar labial lobes are subtriangular and relatively elongated. At the other end of the spectrum, KNM-ER 42842 is a right hemi-mandible with complete tooth row. Taphonomic processes damaged the occlusal surface of the teeth. The molar labial lobes are triangular. The M1 is the only molar with a central pillar, but is weak and

disconnected from the occlusal surface of the tooth. The M3 hypoconulid is small relative to the rest of tooth. The specimens are similar in size and in the degree of tooth wear.

TRAGELAPHUS KYALOI Harris, 1991

Allia Bay material: 5558, Rt. horn core; 19984, frontlet with four additional horn cores fragments; 39447, proximal Lt. horn core fragment; 39449, proximal Rt. horn core fragment; 39451, proximal Lt. horn core fragment; 43341, Rt. M2; 43812, Rt. horn core fragment.

Tragelaphus kyaloi is one of the most abundant bovid species at Kanapoi

(Geraads et al. 2013). The horn cores are of medium to large size and inserted closely to one another on the skull. They are approximately oval in cross-section and slightly compressed anteroposteriorly at the base (basal DAP/DT index = c. 0.78) and they progressively become more triangular at the tips. They have two keels and the right horn core spirals clockwise. Specimens include a complete right horn core (KNM-ER 5558) and frontlet with broken horn cores in many pieces (KNM-ER 19984, Figure 6A).

52

Rt. horn core Lt. horn core

Specimen Prox ap Prox tr Prox ap Prox tr number

ER 19984 46 56 46 56

ER 39446 44.5 31.5 - -

ER 39447 - - 42.5 53.5

ER 39449 33.5 43 - -

ER 39451 - - 35 50.5

ER 39457 24.5 23 - -

ER 40217 29.5 26 - -

ER 43067 - - 40 52

ER 43814 26.5 20* 26 -

ER 43821 - - 23 24.5

ER 43828 - - 33 44

ER 43830 2 1.65 - -

Table 2: Horn Core Measurements (in mm)

TRAGELAPHUS cf. SARAITU Geraads, Melillo and Haile-Selassie, 2009

Allia Bay material: 18891, two nearly complete horn cores with frontlet; 38999, distal Lt. horn core fragment; 39443, Lt. horn core fragment; 39444, distal Lt. horn core fragment;

39445, distal Lt. horn core fragments; 42540, Lt. mandible fragment (P4-M1); 43067, Rt. almost complete horn core.

53

We tentatively assign seven horn cores in the collection to Tragelaphus saraitu.

This species is currently known from penecomtemporaneous sediments from

Woranso-Mille in the Afar Basin (Geraads et al. 2009) as well as from the Mursi

Formation (Drapeau et al. 2014). The horn cores are of relatively small size, strongly spiraled (clockwise for the right horn core), have a triangular cross-section at the base

(basal DAP/DT index = c. 0.8) and are only slightly inclined. They have a strong posterolateral keel and a faint anterolateral keel. These horn cores are also more tightly spiraled than in T. kyaoloi and the latter’s horn cores are also less slender, flatter, and generally larger. Well-preserved specimens include KNM-ER 38999, a complete left horn core that is tightly spiraled. KNM-ER 43067 possesses similar features and is about

75% complete, with only the distal part of the horn core missing (Figure 6B). The specimens are also distinct from Tragelaphus rastafari (Bibi 2011), a relatively newly described species, which comprises the fossil remains previously attributed to T. nakuae and T. aff. nakuae specimens older than 2.98 Ma. T. rastafari is a large species and its horn cores are more slender, more triangular in cross-section and more tightly twisted than the specimens described here. T. rastafari dental remains are not currently extensively described, and therefore it is not possible at the moment to determine if any of the dental remains from Allia Bay overlap with that of this species. Allia Bay specimens attributed to T. saraitu are also larger than T. moroitu, a small-bodied

Tragelaphus species from the Mio-Pliocene (Haile-Selassie et al 2009). T. nkondoensis from Uganda has horn cores that are more upright than any of the Allia Bay tragelaphin specimens.

54

AEPYCEROTINI Gray, 1872

AEPYCEROS Sundeval, 1847

AEPYCEROS aff. SHUNGURAE Gentry, 1985

Allia Bay material: 42541, Rt. P3-M3; 43814, almost complete Rt. horn core.

There are 45 fossils assigned to Aepycerotini (or cf. Aepycerotini). Teeth attributed to Aepycerotini in this collection are similar in shape and size to modern

Aepyceros melampus. The upper teeth are simple with few defining features. They have an oval to square shape with rounded styles and lack a basal pillar. The M3 is posteriorly convex. The lower teeth are more distinctive, with their triangular lobes, weak lingual ribs, and distinctively curved styles. The parastyle is curved, weak and sharp and the enamel is rugose. The M3 has relatively a large third labial lobe. KNM-ER 42541 is a set of connected right teeth, from P3 to M2. Their central enamel islands have the shape of a butterfly (Figure 6O). In addition, they have rounded and pronounced styles and labial ribs and lack pinched lingual lobes.

We tentatively identified four horn cores to Aepycerotini, with only one positively attributed to the tribe. KNM-ER 43814 is an almost complete right horn core, which is mediolaterally compressed and is curved backward. It is somewhat sigmoid in posterior view and has faint transverse grooves.

55

ANTILOPINI Gray, 1821

We identified thirteen remains as Antilopini. The teeth are small and moderately hypsodont. KNM-ER 42833 is a small worn upper molar, with poorly pronounced labial ribs and styles and square lobes (Figure 6N).

Figure 6 (next page): (A) Tragelaphus kyaloi, KNM-ER 19984. 1. Frontlet with four additional horn cores fragments, posterior view. 2. Frontlet, anterior view. (B)

Tragelaphus cf. saraitu, KNM-ER 43067, almost complete Rt. horn core (C)

Tragelaphus sp. (small), KNM-ER 39457. (D) Reduncini, KNM-ER 39001, Rt. horn core fragment. (E) cf. Gazella, KNM-ER 39002, Lt. horn core fragment. (F)

Alcelaphini, KNM-ER 40217, distal Rt. horn core fragment. (G) Hippotragini,

KNM-ER 39446, proximal Rt. horn core with partial frontal bone. (H)

“Neotragini”, KNM-ER 42976, Rt. M3. (I) Simatherium kohllarseni, KNM-ER

43437, Rt. M3. (J) Tragelaphus sp., KNM-ER 42687, Rt. hemi-mandible with complete tooth row. (K) Tragelaphus, KNM-ER 43394, Lt. M3. (L) cf. Parmularius,

KNM-ER 42879, Rt. M2.. (M) Alcelaphini, KNM-ER 43344, Rt. M2. (N) Antilopini,

KNM-ER 42833, Lt. M1 or 2. (O) Aepyceros aff. shungurae, KNM-ER 42541, Rt. P3-

M3. Scale bar = 5 cm for A-G and 1 cm for H-O.

56

57

GAZELLA de Blainville, 1816 cf. GAZELLA

Allia Bay material: 39002, Lt. horn core fragment; 42670, Rt. M1.

Two remains are described as either Gazella sp. or cf. Gazella. KNM-ER 39002 is a left horn core fragment (Figure 6E). With its rounded cross-section and slight groove, it resembles Gazella sp. specimen KNM-LT 23653 in size and shape. A small sized right upper first molar is also assigned to cf. Gazella.

ALCELAPHINI Brooke in Wallace, 1876

We classified a total of 44 fossil specimens as members of the tribe Alcelaphini.

Alcelaphins are the most hypsodont bovids in the collection. The occlusal surface of the teeth is small and rounded in the lower teeth and more square in the upper teeth. The molars have large bases compared to their occlusal surfaces (Figure 6M). The central enamel islands are large relative to the occlusal surface of the tooth and are shaped like a butterfly. The labial and lingual ribs are rounded. Some specimens display a weak basal pillar, but most do not. Some lower molars possess weak goat folds. We did not identify the vast majority of the remains attributed to this tribe to a genus. However, two distinct alcelaphin morphotypes are present in the assemblage: a small one, which could be

Damalborea and a large one, possibly Parmularius sp.

58

cf. PARMULARIUS Hopwood, 1934

2 1 2 Allia Bay material: 42668, Rt. M ; 42880, Rt. M1; 42888, Lt. M or M ; 43233, Lt. M2.

Specimens include four molars, but no horn cores have been attributed to

Parmularius so it is difficult to confirm a taxonomic attribution. KNM-ER 42879 is a particularly well-preserved right M2 (Figure 6L), which bears similarities to KNM-KP

109 from Kanapoi, identified as Alcelaphini by Geraads and colleagues (2013). The tooth is robust and hypsodont, has strong labial ribs and lacks a central pillar. The central enamel island is butterfly-shaped.

cf. DAMALBOREA gen. nov.

Allia Bay material: 39461, horn core fragment; 40217, distal Rt. horn core fragment;

43231, Lt. M3.

These small sized alcelaphins have affinities to both Damalborea and Damalacra.

Notably, KNM-ER 40217 (Figure 6F) is a right distal horn core fragment, which closely resembles KNM-KP 71 from Kanapoi, identified as ?Damalborea (Geraads et al. 2013).

The Allia Bay and Kanapoi specimens are comparable in their perfectly rounded cross-section, moderately marked transverse ridges, the fact that they are upright and their lateral flattening. One of their most distinctive features of this specimen is the way the base of the horn core flares into the skull without any visible pedicel. KNM-ER

39461 is a small horn core fragment with similar features.

59

BOVINI Gray, 1821

Within the Allia Bay collection, Bovini are very rare. We identified four specimens to this tribe, described in details below.

SIMATHERIUM KOHLLARSENI Dietrich, 1941

Allia Bay material: 230, calvaria with both horn cores and partial face; 2954, Lt. upper

M; 43437, Rt. M3; 43438, Rt. M3.

KNM-ER 43438 is the largest bovid tooth in the assemblage. The right upper third molar was identified in a previous publication as Simatherium sp. based on its large size and pinched lingual lobes (Harris 1991). It has pronounced rounded labial ribs and a large butterfly-shaped interior enamel island. On the distal lingual lobe, there is a small dent in the outline of the central enamel island. This unique featured is shared with similarly sized and shaped KNM-KP 29265, an upper second molar specimen from

Kanapoi identified as Simatherium sp. However, the latter has a labial pillar, and the

Allia Bay specimen does not. The upper molars KNM-ER 2954 and KNM-ER 43437 are also assigned to Simatherium sp. (Figure 6I). KNM-ER 230 is a calvaria with both horn cores and a partial face. It is identified in a previous publication as Simatherium cf. kohllarseni (Harris 1991, fig. 5.15), but was collected in the 1960s and the horizon from which it originates is unknown (Harris 1991).

60

HIPPOTRAGINI Sundevall, 1845

HIPPOTRAGUS Sundevall, 1845

Allia Bay material: 39446, proximal Rt. horn core with partial frontal bone; 39458, Rt. horn core fragment.

Two horn cores have been attributed or tentatively assigned to Hippotragini.

KNM-ER 39446 is the proximal half of a right horn core with part of the frontal bone, including the supraorbital foramen (Figure 6G). We tentatively assigned it to

Hippotragus aff. gigas. The well-preserved frontal bone of KNM-ER 39446 shows that the horns were close together and inserted closely above the orbit. The horn core is curved backwards about 45 degrees and slightly laterally divergent in anterior view. Oval in cross-section, it has shallow and faint longitudinal grooves and a single more marked groove placed posteriorly. It is less strongly curved than in the Kanapoi specimen

KNM-KP 36604 identified as cf. Tchadotragus. KNM-ER 39446 shares many morphological traits with KNM-WT 17486, identified as Hippotragus gigas. The right horn core fragment KNM-ER 39458 is attributed to cf. Hippotragini because despite its lack of diagnostic features it shares a number of traits with Hippotragini KNM-LT 38435, including their size, an ovoid cross-section, longitudinal grooves and a similar curvature.

REDUNCINI Knottnerus-Meyer, 1907

Within the Allia Bay collection, we classify 14 fossils as reduncins, about half of which are horn cores. Two species have been identified thus far in the collection: Kobus laticornis and oricornis.

61

Amongst the dental specimens, the best preserved is KNM-ER 30351, a mandible

with the right P4 to M3 in place, which displays goat folds and a central pillars connected to the occlusal surface of the tooth. The teeth resemble that of the modern Kobus ellipsiprymnus (common waterbuck) in shape and size. KNM-ER 40200 is a right mandible fragment with a M2 in place. The specimen is taphonomically altered. The molar possesses a small goat fold and pronounced lingual ribs, so we identify it at least tentatively as a reduncin. Other dental specimens include molars KNM-ER 43304 and

42646. KNM-ER 43052 is a horn core tip, which could belong to K. presigmoidalis due to its shared features with comparative specimen KNM-LT 189. It displays very deep transverse grooves, and its end is wide and rounded, but it is too fragmentary for a clear attribution. KNM-ER 39001 is a right distal horn core fragment that is slender and upright, with a flattened lateral side (Figure 6D). Its one deep groove with additional fainter grooves resembles KNM-KP 29264, identified as ?Kobus by Geraads and colleagues in 2013. The cross-sections of the two specimens are similarly rounded.

KOBUS LATICORNIS Harris, 2003

Allia Bay material: 39460, Lt. horn core fragment.

We attribute one specimen to the species Kobus laticornis. KNM-ER 39460 is a left horn core fragment with transverse ridges, a faint keel, and a relatively flat oval cross-section. The lateral ridges are more pronounced than the medial ridges. It curves slightly backward. This specimen is more upright and slender than K. presigmoidalis specimens such as KNM-LT 189.

62

KOBUS ORICORNIS Gentry, 1985

Allia Bay material: 5026, proximal horn core fragment; 17578, Rt. proximal horn core fragment with frontlet.

We attribute two proximal horn core fragments to Kobus oricornis: KNM-ER

5026 and 17578. They are long, upright and slender specimens with transverse ridges.

“NEOTRAGINI” Sclater and Thomas, 1894

Although the monophyletic nature of the tribe Neotragini is contested in both the morphometric and genetic literature (Bärmann and Schikora 2014; Bibi et al. 2009;

Hassanin et al. 2012), we use this category here to refer to an ecologically cohesive group of small sized antelopes that cannot be classified as Antilopini or Aepycerotini.

Neotragins are traditionally understood as small-sized bovids that consume a high proportion of dicots (Gagnon and Chew 2000). We only classified four specimens as neotragins. KNM-ER 42842 is a very small sized and almost complete lower third molar.

KNM-ER 42976 is a right M3 that resembles Lothagam bovid KNM-LT 176 (Figure 6H).

The Allia Bay specimen is smaller than the remains from Kanapoi attributed to

Raphicerus. It has triangularly shaped labial lobes. The third (posterior) labial lobe is small and without a central valley, which justifies its inclusion to Neotragini. We tentatively attribute it to the genus Madoqua. KNM-ER 40172 is a small left mandible fragment with five very worn teeth in sockets. KNM-ER 40173 is similar

63

Specimen number MRL M1L M1W M2L M2W M3L M3W

ER 30345 6.05 1.35* - 2* 1.35 3* 1.3

ER 30351 - - - 1.6 1.95 2.5 1.05

ER 40173 - - - 0.85 0.4 1.2 0.45

ER 40174 - 0.9 0.55 1.1 0.85 - -

ER 40200 - - - 2.3 1.35 - -

ER 40201 - 1.5 1 1.9 1.2 - 1.15

ER 42540 - 1.8 1.15 - - - -

ER 42542 - - - 1.95 0.65 - -

ER 42546 - - - 1.60 0.85 - -

ER 42551 - - - 1.62 1.05 - -

ER 42556 - - - 1.67 0.74 - -

ER 42559 - - - 2.25 1.05 - -

ER 42600 - - - 1.74 0.94 - -

ER 42601 - - - 1.85 1.00 - -

ER 42602 - 1.95 0.90 - - - -

ER 42604 - 1.50 1.10 - - - -

ER 42607 - - - - - 2.60 1.05

ER 42608 - - - 2.00 1.05 - -

ER 42612 - 1.80 1.10 - - - -

ER 42654 - - - 1.53 0.70 - -

ER 42655 - 1.5 1.3 - - - -

ER 42660 - - - 1.65 1.10 - -

ER 42664 - - - 2.05 1.20 - -

ER 42668 - - - 1.79 1.04 - -

ER 42673 - - - - - 3.25 1.30

64

ER 42676 - - - 1.90 0.85 - -

ER 42677 - 1.75 1.2 2.05 1.2 - -

ER 42685 - 1.85 1.00 - -

ER 42687 6.75 1.55 1.7 2.1 1.35 2.95 1.35

ER 42694 - 1.85 1.15 - - - -

ER 42695 - - - 2.00 1.15 - -

ER 42697 - 1.85 0.95 - - - -

ER 42699 - - - 1.71 0.82 - -

ER 42812 - - - 0.74 1.39 - -

ER 42823 - 1.6 1.2 1.9 1.25 - -

ER 42829 - - - - - 2.20 0.70

ER 42831 - 1.90 1.10 - -

ER 42841 6.2 1.4 1.05 1.8 1.1 3.1 1.1

ER 42842 6.8 1.7 1.1 1.95 1.25 2.8 1.1

ER 42880 - 1.67 0.97 - - - -

ER 42882 - - - 2.23 1.05

ER 42883 - - - 2.20 0.976* - -

ER 42884 - - - 2.08 0.96 - -

ER 42898 - - - 1.52 0.85 - -

ER 42944 - - - 1.95 1.05 - -

ER 42946 - - - 1.60 0.80 - -

ER 42951 - - - 1.75 0.95 - -

ER 42953 - - - 1.70 1.05 - -

ER 42955 - - - - - 2.60 0.95

ER 42956 - - - - - 2.80 1.10

ER 42957 - - - - - 2.85 0.95

ER 42958 - - - 1.70 0.90 - -

ER 42959 - - - 1.90 0.90 - -

65

ER 42960 - - - 1.75 0.90 - -

ER 42961 - - - 2.00 1.10 - -

ER 42976 - - - - - 1.23 0.46

ER 43230 - - - - - 2.74 1.06

ER 43233 - - - 2.4 1.1 - -

ER 43234 - - - 1.94 0.84 - -

ER 43342 - - - 2.08 1.09 - -

ER 43354 - 1.20 1.99 - - - -

ER 43356 - - - 1.94 1.25 - -

ER 43362 - - - 2.10 1.20 - -

ER 43363 - - - 1.60 0.72 - -

ER 43369 - - - 0.99 2.25 - -

ER 43373 - - - - - 3.38 1.39

ER 43386 - 1.74 1.20 - - - -

ER 43391 - - - 1.65 0.78 - -

ER 43395 - - - - - 2.93 0.95

ER 43396 - - - - - 1.86 0.77

ER 43439 - - - 1.83 0.94 - -

LOWER TEETH: Measurements: MRL molar row length; M#L molar number length; M#W molar number width; * value is an estimate.

Specimen number MRL M1L M1W M2L M2W M3L M3W

ER 17370 - - - 2.24 1.25 - -

ER 42541 - 1.45 1.4 2 1.6 - -

ER 42543 - 1.75* 1.9 - - - -

ER 42547 - 1.29 1.08 - - - -

ER 42557 - - - - - 1.95 1.3

66

ER 42605 - - - 1.70 1.75 - -

ER 42609 - - - - - 2.05 1.75

ER 42611 - - - 1.95 1.44 - -

ER 42646 - - - 2.35 1.20 - -

ER 42653 - 1.70 1.60 - - - -

ER 42657 - 1.35 1.65 - - - -

ER 42658 - - - - - 1.80 1.15

ER 42659 - - - - - 1.66 1.21

ER 42667 - - - 2.00 1.65 - -

ER 42670 - 1.17 0.67 - - - -

ER 42681 - - - 1.85 1.80 - -

ER 42682 - - - 1.95 1.4* - -

ER 42686 - 1.60 1.70 - - - -

ER 42688 - 1.45 1.7 2 1.8 2.2 1.9

ER 42811 - - - - - 1.90 1.55

ER 42812 - - - - - 2.20 1.45

ER 42814 - - - 1.89 1.41 - -

ER 42815 - 1.8* 1.8 - - - -

ER 42819 - - - - - 2.20 1.45

ER 42822 - - - - - 2.4 -

ER 42879 - - - 1.61 2.37 - -

ER 42885 - - - - - 2.03 1.40

ER 42886 - - - 2.18 1.66 - -

ER 42891 - - - - - 1.49 2.21

ER 42900 - - - 1.75 1.2 - -

ER 42966 - - - 2.1* 1.65 - -

ER 42962 - - - - - 2.45 1.80

ER 43023 - - - - - 2.20 1.70

67

ER 43231 - - - - - 1.95 1.04

ER 43239 - 1.55 1.18 - - - -

ER 43300 - 1.90 1.18 - - - -

ER 43301 - - - - - 1.93 1.88

ER 43302 - 1.66 1.47 - - - -

ER 43305 - - - - - 2.46 1.61

ER 43339 - 2.2 1.5 - - - -

ER 43341 - - - 2.33 1.80 - -

ER 43344 - - - 1.99 1.47 - -

ER 43346 - - - 1.89 1.62 - -

ER 43355 - - - - - 2.46 1.61

ER 43357 - 1.70 1.70 - - - -

ER 43361 - - - - - 2.43 1.89

ER 43365 - 2.09 1.47 - - - -

ER 43366 - - - 2.43 1.89 - -

ER 43368 - - - - - 2.47 1.56

ER 43372 - 1.38 1.06 - - - -

ER 43375 - 1.98 1.52 - - - -

ER 43384 - - - 1.97 1.85 - -

ER 43394 - - - - - 2.65 2.07

ER 43397 - 1.64 1.66 - - - -

ER 43438 - - - 2.92 2.27 - -

UPPER TEETH: Measurements: MRL molar row length; M#L molar number length; M#W molar number width; * value is an estimate.

Table 3: Measurements (in mm) for all Bovidae molars in the Allia Bay assemblage Postcranial ecomorphology

68

morphologically, but retains a complete M2 and M3. The third lower molar has labial lobes that are the same size, a distinctive neotragin feature. We found no horn core remains from this tribe, which may be due to differential preservation, as these are small sized animals.

Postcranial ecomorphology examines the relationship between morphology and function in a given environment. In bovids, ecomorphology methods are employed on crania and mandibles (Forrest et al. 2018; Solounias and Moelleken 1993; Spencer 1997) and, most frequently, on postcrania. In bovids and other cursorial mammals, differences in post-cranial morphology are linked to predator avoidance strategies that reflect habitat structure (Kappelman 1988; Plummer et al. 2008). This method has the advantage of not relying on taxonomic uniformitarianism, but the disadvantage of relying on a larger sample of bones than is sometimes available in fossil collections from the Pliocene, which can be highly fragmented.

Although reconstructing environments has been achieved successfully using other postcranial remains (Kovarovic and Andrews 2007; Plummer and Bishop 1994), this study is limited to the astragali because of small sample sizes for other elements.

Astragali are compact and robust, preferentially well preserved and their shape makes them easily recognizable (Plummer et al. 2008). Astragali are also the most widely studied bones for ecomorphological purposes (Plummer et al. 2008). Bovids that live in heavy cover environments have anterior-posteriorly expanded astragali, which allow them to move between trees in complex settings, whereas more cursorial bovids living in open settings tend to have superior-inferiorly compressed astragali (DeGusta and Vrba

69

2003).

This method has been employed to reconstruct the paleoenvironments at various hominin-bearing sites including the Shungura Formation (Barr 2015; Plummer et al.

2015), Olduvai Gorge (Plummer and Bishop 1994) and Laetoli (Bishop et al. 2011;

Kovarovic and Andrews 2007). To date, no ecomorphological analyses have been attempted for assemblages from Allia Bay, Kanapoi or Mursi.

Carbon and oxygen isotopes of mammalian tooth enamel

Enamel carbon isotopic data is commonly used as a dietary indicator in ecological and paleontological studies. This method is based on the fact that there exists three different pathways in terrestrial tropical plants photosynthesis. Warm-growing-season and low-elevation grasses and sedges use the C4 photosynthetic pathway for carbon fixation.

Those that use the C3 pathway include a much broader array of vegetation: trees, shrubs, bushes, sedges and cold-growing-season grasses. In tropical environments, plants that use

13 the C3 photosynthetic pathway yield values that are C-deplete compared to those that use the C4 photosynthetic pathway (respectively, C3 plants around -22 to – 35 per mil

(‰) compared to C4 plants at about -10 to -15‰) (Cerling et al. 2015; Koch 1998; Lee-

Thorp et al. 1989). Crassulacean acid metabolism (CAM) is reserved to succulents, thus generally limited to desert environments (Lüttge 2004).

Carbon isotope-based paleodietary studies are concerned with the types of plants

(i.e., C3 or C4) ingested by herbivores. The enamel bioapatite in herbivore teeth reflects

70

the carbon isotopic composition of their diet. Herbivores can be considered grazers (G)

13 when δ C values preserved in their enamel are -1‰ or higher, browsers (B) when the values are -8‰ or lower, and mixed feeders (M) when the value is in between (Cerling et al. 2015). Mixed feeders are herbivores that are non-specialists (eurybiomic) and can thus more easily adapt to more variable environments and thrive in highly seasonal ecosystems (Cantalapiedra et al. 2014). For example, modern African impalas have an extremely flexible diet, which can vary from pure browse to exclusively grazing according to the competition, the seasons or the local habitat conditions (Codron et al.

2006; Pérez-Barbería et al. 2001). The diet of a species can also vary according to its feeding strategy (non-selective bulk or selective feeding), which itself depends on many factors including availability of food types, time allotted to feeding and biological requirements like gastrointestinal morphologies and capacities (Beekman and Prins

1989). Animals may prefer a certain type of food, but their diet may in fact be impacted by another type of food. This distinction between preferred vs. fallback food (Marshall and Wrangham 2007)is thus central to studying isotopically the diet of mixed feeders.

Because of the adaptability of C3-C4 mixed feeders, efforts were made to focus our sampling on taxa known to have a more flexible diet. We used the notation G:M:B to characterize the ratio of grazers to mixed feeders to browsers in the assemblage (Manthi et al. 2018).

The oxygen isotopic composition of mammalian enamel bioapatite is another indicator of paleoecology, but environmental interpretations are complex, with a number of biotic and abiotic factors involved during the incorporation of oxygen environmental water into the enamel structure (Luz et al. 1984). Factors influencing the composition of

71

oxygen in water include temperature, altitude and distance from the oceanic source region, i.e. “continental effect” (Krishnamurthy and Bhattacharya 1991). Within animals oxygen isotope ratios reflect body water, which is itself linked to physiology, drinking behavior and leaf water (Koch 1998; Sponheimer and Lee-Thorp 1999). In obligate drinkers, which are animals that drink water directly from a meteoric source water, the tooth enamel reflects the oxygen isotopic composition of the water source which may reflect that of rainfall contributing to that source, modified to some degree by evaporation

(Dansgaard 1964). Non-obligate drinkers, or arid-adapted taxa, can survive without direct access to a body of water and instead obtain water from the plants (leaves, fruits, or other parts) they ingest (Ayliffe and Chivas 1990), which are 10-30‰ enriched compared to the meteoric source water (Yakir 1998). The oxygen isotopic composition of their enamel will thus be enriched compared to that of rainfall and influenced by the type and part of the plant they ingest. Since the isotopic composition of local source water is influenced by rainfall and other local climatic factors, geographic locations or periods are not directly comparable (Koch 1998).

Dental mesowear and hypsodonty

Macroscopic dental wear facets (i.e., mesowear) created on the molars of ungulates during their lifetime vary in size and shape according to the properties of the foods ingested (abrasion) and the tooth-on-tooth wear during mastication (attrition) (Fortelius and Solounias 2000). Studies indicate that the shapes of the facets differ between grazing

(grass-eating) and browsing (leaf-eating) taxa. Animals that feed on leaves possess teeth

72

with higher and sharper cusps than those that feed on grass.

Similarly, the relative height of the tooth crown, or hypsodonty, has been shown to correlate with diet (Janis 1988; Williams and Kay 2001). The simplest explanation for differences in hypsodonty is that grazing herbivores have higher tooth crown than browsers. In addition, more recent research have demonstrated that factors such as habitat type and rainfall are also correlated with hypsodonty and that, in fact, all these factors are interrelated (Fortelius et al. 2006; Mendoza and Palmqvist 2008). Hypsodonty is a genetic trait and is related to long-term evolution (Damuth and Janis 2011), whereas mesowear reflects diet adopted during an animal’s life. Thus, although both mesowear and hyposodonty provide information about dietary ecology, they do so on different scales. These measures can be combined to improve the accuracy of paleoenvironmental reconstructions (Andrews and Hixson 2014; Damuth and Janis 2011; Fortelius and

Solounias 2000; Fraser and Theodor 2011).

Mesowear is commonly used to analyze faunal remains at hominin sites including Laetoli (Kaiser 2011), Kanapoi (Dumouchel and Bobe In review) as well as the Kibish (Rowan et al. 2015) and Shungura Formations (Blondel et al. 2018).

Methods

Taxonomy

We follow Vrba’s approach in using bovid tribal proportions and analogies with their extant relatives to generate hypotheses about the paleoenvironments (Vrba 1980). Indices

73

of richness and diversity are also calculated for the three collections. Significant differences in these indices would suggest that the communities are drawn from different habitat types. Ecological indices of richness and diversity tend to be higher in more forested environments, although it can vary substantially according to multiple ecological factors (Andrews 1996; Magurran 1988). Taxa are classified in dietary categories as follows: grazers (Alcelaphini and Reduncini), mixed feeders (Aepycerotini,

Antilopini, Bovini, Hippotragini), browsers (Neotragini, Tragelaphini) (Cerling et al.

2013).

Postcranial ecomorphology

We collected standardized linear measurements for post-cranial ecomorphology on bovid astragali from the Allia Bay (n=18), Kanapoi (n=34) and Mursi (n=1) assemblages (Barr

2014; Barr 2015). Both left and right 90-100% complete bones were used. We analyzed data from Allia Bay and Kanapoi astragali using Discriminant Function Analyses following Barr and Scott (2014), which classifies each specimen in a discrete habitat category; Forest (or Closed), Heavy Cover, Light Cover and Open.

The Mursi assemblage only contains one complete bovid astragalus (MR-617), thus results were not included in statistical analyses.

Dental mesowear and hypsodonty

We evaluated complete or nearly complete adult molars in the three assemblages

74

according to stages of hypsodonty, occlusal relief, and cusp shape.

As recommended, second molars were used for mesowear analyses (Fortelius and

Solounias 2000). Since the publication of the original mesowear method in 2000, additional support for the use of the M2 in mesowear studies has been shown (Louys et al. 2012; Louys et al. 2011). We used isolated teeth as well as teeth still in place in the jaw. Only lightly worn teeth were used. We used the “mesowear stage” method, an extension of the original mesowear method by Fortelius and Solounias (2000), to score the teeth (Kaiser et al. 2013). Each specimen was scored in buccal (lateral) view according to 1) cusp shape (sharp, rounded or blunt) and 2) occlusal relief (high or low cusps) (Figure 3). We then converted the traits into mesowear stages from 0 (hyper browser) to 4 (hyper grazer). The scores thus represent the combined effect of abrasion and attrition. One of us (Dumouchel) performed all the scoring to control for inter-observer bias. Intra-observer bias was found to be negligible (Kaiser et al. 2000).

Hypsodonty (Hyp) was evaluated by dividing the height of the tooth by its buccolingual width (Figure 3, (Janis 1988)). Only lightly worn or unworn M3 were used

(stage 0 or 1, according to a scoring scale of 0 (unworn) to 3 (very worn)). Some teeth are from the same individuals as the teeth used for the mesowear analysis, but most are not.

Next, the tribe or genus average was computed and classified as one of three possible categories 1) mesodont (Hyp = 1.5-3), 2) brachydont (Hyp < 1.5) and 3) hypsodont (Hyp

> 3) (Damuth and Janis 2011).

Mesowear sample sizes are as follows: Allia Bay (n=67), Kanapoi (n=65) and

Mursi (n=1). Values from a total of 46 individuals were used to compute tribe averages and come from the assemblages as follows: Allia Bay (n=16), Kanapoi (n=28) and Mursi

75

(n=2).

Dental enamel carbon and oxygen stable isotope values

We used a high-speed rotary drill fitted with a small diamond bit to clean the surface of each tooth sampled before collecting approximately 10 to 15 mg of enamel powder. The drill was cleaned between each sample to avoid contamination. The enamel powders were imported to the United States following the guidelines of the National Museums of

Kenya and Ethiopia.

We treated the samples of enamel powder with 1 M acetic acid-calcium acetate buffer to remove any secondary carbonates and organic matter. We performed a series of rinses with deionized water (Koch 1998). Samples were dried at 25 °C for 24h. Samples were reacted with 103% phosphoric acid and analyzed in a ThermoFisher Scientific Delta

V Advantage Isotope Ratio Mass Spectrometer at the stable isotope laboratory at the

University of South Florida. The precision of replicate analyses of these standards within a given run was less than 0.1‰. We report the results using the standard per mil (‰) notation:

13 18 3 δ C or δ O = [Rsample/Rstandard - 1] x 10 ‰

13 12 18 16 where Rsample is the C/ C or O/ O ratios of the sample and Rstandard is the ratio of the reference standard. δ13C and δ18O values of tooth enamel are reported on the VPDB scale, which was normalised using replicate analyses of IAEA and internal reference

76

materials NBS-18 (δ13C= -5.01‰ and δ18O=-23.01‰) and NewCar (δ13C= 4.04‰ and

δ18O=-3.38‰).

We examined a total of 41 carbon and oxygen stable isotope values from bovid teeth from Allia Bay. Data were obtained from new (n=22) and published data (n=19)

(Cerling et al. 2015; Schoeninger et al. 2003). Comparative data also come from new

(Mursi, n=7) and published (Kanapoi: n=27; Mursi: n=3) sources (Cerling et al. 2015;

Drapeau et al. 2014; Manthi et al. 2018). The specimens selected were identifiable to at least the tribe level and late-erupting teeth were preferred. The distribution of the 78 new and published specimens is as follows: Aepycerotini (n=22), Alcelaphini (n=14),

Antilopini (n=4), Bovini (n=2), Ovibovini (n=1), Reduncini (n=1) and Tragelaphini (n=

34).

All statistical analyses were performed in R 3.4.3.

Results

Taxonomic analysis

Tragelaphins dominate the Bovidae at both Allia Bay and Kanapoi, but they are more abundant at Allia Bay (46.5% vs. 37.6%) (Figure 7). The Mursi collection only comprises three bovid tribes, and is dominated by remains attributed to Aepycerotini

(43.64%). The second most abundant tribe is Alcelaphini at Allia Bay (17.25%),

Aepycerotini at Kanapoi (19.21%) and Tragelaphini at Mursi (14.55%). Bovids classified

77

as browsers are the most abundant in the Allia Bay assemblage (54.41%). The dietary category is also well represented in the Kanapoi assemblage (47.48%), and less so in the

Mursi collection (25.7%), which is contrary to our predictions. The Mursi assemblage had the most mixed feeders (68.6% vs. 23.72% for Allia Bay and 33.32% for Kanapoi), and the least grazers (5.71% vs. 21.95% for Allia Bay and 19.22% for Kanapoi). Despite substantial overlap between the faunal abundance at Allia Bay and Kanapoi, the tribe distribution is significantly different between the two faunal assemblages (p= 0.0002 with

Monte Carlo Fisher’s exact test with 5000 simulations). One of the noteworthy differences is that small sized Neotragini and Antilopini are more common at Kanapoi than Allia Bay. In addition, Bovini is proportionally much more abundant at Kanapoi than at Allia Bay. The Mursi assemblage is characterized by a low Simpson’s index

(0.475), which is most likely a consequence of the small sample size (n=55).

Locomotor strategies

The ecomorphological data shows an environment dominated by open settings at

Kanapoi (61.3%), but a more mixed environment at Allia Bay, with each of the four habitat types represented by values between 46.7% and 13.3% (Figure 8). The difference between the two assemblages is however not statistically significant (p= 0.4335 with

Monte Carlo Fisher’s exact Test with 5000 simulations). The only complete bovid astragalus from the Mursi collection (MR-617) is small and its analysis suggests an adaptation to a low cover environment. A larger sample size is needed before well-supported ecomorphology-based hypotheses can be formulated about the Mursi

78

environments.

Figure 7: Comparison of the bovid composition between the three collections, excluding indeterminate.

Dietary Ecology

Tragelaphini

Tragelaphins are the dominant bovid tribe in c. 4 Ma assemblages in eastern African such as the ones mentionned in this paper, Asa Issie (White et al. 2006) and Woranso-Mille

(Geraads et al. 2009). This is also the case at Allia Bay. Mesowear analyses present Allia

Bay tragelaphins as mixed feeders (1.8 ± 1.4), as is expected for this tribe (Greenacre and

79

Vrba 1984) (Figure 9). They incorporate more graze into their diet than their Kanapoi counterparts (1.4 ± 1.12). The single specimen from Mursi that could be analyzed has a mesowear score of zero, the value associated with hyper browsers. Tragelaphins were also browsing at Aramis (0.5) and Woranso-Mille (Curran and Haile-Selassie 2016;

White et al. 2009a). The hypsodonty index of the tribe at Allia Bay is 2.21, categorizing them as mesodonts. Hyposodonty indices were lower at both Kanapoi (1.37) and Mursi

(1.71). These values are consistent with a more attrition-dominated diet.

Figure 8: Stacked bar plots showing the relative proportions of habitat predictions based on astragalar ecomorphology for the assemblages from Allia Bay and Kanapoi.

80

Figure 9: Mean hypsodonty against mesowear score by tribe in the three fossil collections. Black diamonds indicate the mean. Only tribes with both hyposodonty and mesowear values are included in the graph, n= 156.

81

In Allia Bay tragelaphins, the mean carbon isotope is consistent with browse-dominated mixed feeding: -7.77 ± 1.61 per mil units (‰) (n=21, Figure 10A and

10B). Individual values vary between -10.1 and -4.6 ‰ and the G:M:B ratio is 0:11:10.

Perhaps because of the relatively large sample size of the tribe, the distribution shows many specimens considered outliers. This phenomenon underscores the importance of increasing sample sizes when possible. This variation in the signature in Tragelaphini is also seen in Mursi (n=5) and Kanapoi (n=8) specimens. The mean carbon isotope value associated with tragelaphins from the Mursi assemblage is more negative (-13.18 ± 0.97

‰, with a G:M:B ratio of 0:0:5) than at Allia Bay, indicating that the individuals incorporated larger proportions of browse in their diet. On the contrary, individuals from this tribe had slightly less depleted but overlapping values at Kanapoi (-7.04 ± 3.32 ‰,

G:M:B=1:3:4). Oxygen isotope results show that tragelaphines were likely obligate drinkers at Mursi (-2.88 ± 1.64‰), whereas values were more enriched at Allia Bay (0.01

± 1.44 ‰) and Kanapoi (1.01 ± 1.73 ‰) (Figure 10C), indicating water was also obtained from other sources.

Aepycerotini

Aepycerotini is the second most common tribe at Allia Bay, and the dominant tribe at the localities of Aralee Issie and Mesgid Dora from the site of Woranso-Mille (Curran and

Haile-Selassie 2016). They show cursorial adaptations and high tooth crowns, both associated with open habitats (Bobe and Behrensmeyer 2004), but they have an eclectic diet (Bobe and Eck 2001). Aepycertotin molars from the Allia Bay fossil collection are consistent with the mesowear signal of mixed feeders. With a values of 1.9 ± 1.6, their

82

score is intermediate and shows important within-tribe variation (Figure 9). The Allia

Bay Aepycerotini hypsodonty score (1.36) is intermediate relative to the other bovids from this collection. The tribe shows a more grazing signal at Kanapoi, with an average mesowear score of 3.36 ± 0.84, a G:M:B ratio of 2:6:1 and a hypsodonty score of 2.41.

We suggest that the important difference between the hypsodonty scores observed for the

Allia Bay and Kanapoi aepycerotins is attributable to the use of small sample sizes.

The average δ C13 value for the tribe at Allia Bay is -4.62 ± 4.58 ‰ (Figure 10A and 10B), which is consistent with a mixed diet. Similarly to tragelaphins, the results shows important within-tribe variation, with δC13 values varying from 1.1 to -10.1 ‰

(G:M:B = 2:4:2). The diets of the Aepyceros show important differences between the three assemblages analyzed. Their diet incorporates more graze at Kanapoi (-3.08 ±

3.91‰, n=9). Published values are similar for this tribe at the site of Aramis (White et al.

2009a). However, aepycerotins were clearly browsing at Mursi (-9.67 ± 3.2‰,

G:M:B=0:1:4). The Allia Bay bovids fall in the middle of the distribution, but closer to the values for Kanapoi. Oxygen isotope data indicate individuals from Allia Bay (0.55±

1.64‰) and Kanapoi (0.82± 2.48‰) were neither water-dependent nor arid-adapted.

Those from Mursi were likely water dependent (-1.76± 2.27‰). Overall, values obtained from isotope analyses indicating high adaptability and versatility of the tribe for that period and location (Figure 10C).

Antilopini

Allia Bay Antilopini have a mesowear score of 3.0±1, which is a graze-dominated mixed-feeding signal. No specimen was suitable for the evaluation of hypsodonty. At

83

Kanapoi, antilopins display a mesowear pattern similar to the one from their counterparts at Allia Bay (2.56 ± 1.24, Figure 9). The hypsodonty signal of 1.44 is moderate.

The diet of antilopins shows more grazing at Kanapoi than Allia Bay using stable isotopes data. However, the sample sizes are very small: data are available for two individuals from each assemblage. The Kanapoi result is -7.65 ± 1.34‰ and the Allia

Bay result is -9.4 ± 0.83‰ (Figure 10A and 10B). Both signals are close, but the tribe is classified as mix feeding at the former (G:M:B = 0:1:1) and as browsing at the latter

(G:M:B = 0:0:2). Kanapoi Antilopini had a mean oxygen value of 6 ± 0.14‰, which is very enriched. This indicates that it is likely that these bovids were getting their water supply from their diet. A lower, but still relatively enriched value characterizes the tribe at Allia Bay, 3.52 ± 1.74‰ (Figure 10C). No Antilopini remains currently exist in the

Mursi collection.

Alcelaphini

Modern Alcelaphins are grazers, with high tooth crowns (Greenacre and Vrba 1984).

Fittingly, the alcelaphins from the Allia Bay collection display the highest hypsodonty index in the Allia Bay bovid assemblage, with a score of 2.33 (Figure 9). Kanapoi alcelaphins display a similar value (HYP=2.25). Despite these relatively high score, alcelaphins in our datasets are classified as mesodonts. Their mesowear score of 1.7 ±

1.5 typically reflect browse-dominated mixed feeders, suggesting their diet at Allia Bay was not dominated by grazing. Similarly, with an average of 1.3 ± 1.4, the alcelaphin fossils from the Kanapoi collection are interpreted as having a mixed diet, which incorporates some browse. This is similar to the values seen in alcelaphins from

84

Woranso-Mille (Curran and Haile-Selassie 2016).

Within the Allia Bay assemblage, the δC13 tribe average is -1.54 ± 2‰, which is

consistent with a graze-dominated mixed C3-C4 herbivore diet (Figure 10A and 10B, n=

8), but individuals vary from 1.7 to -9.5 ‰ (G:M:B = 3:4:1). These results indicate that members of this tribe had flexible and/or seasonal diets, or that multiple species were represented. Alcelaphins at Kanapoi (n=6) are described by a value of 0.28 ± 3.17‰ and are more typical grazers (G:M:B = 5:1:0). Allia Bay alcelaphins have an average enamel oxygen isotope value of 2.19 ± 1.3‰, which is relatively large and indicates that members of this tribe were arid-adapted taxa. An even more enriched δO18 value characterizes the individuals from Kanapoi (3.72 ± 2.1‰) (Figure 10C). It is unlikely that

Alcelaphini were living at Mursi (Drapeau et al. 2014).

Bovini

Within the Allia Bay collection, only one Bovini molar met the criteria for the mesowear analysis and none met the criteria for the calculation of hypsodonty. The mesowear score for this specimen is 3, a value associated with a grazing diet. This individual has the highest signal within the Allia Bay collection. The individuals from the Kanapoi collection had an average score of 1 ± 1.1, which is in theory consistent with a browsing diet. However, these individuals tended to have rounded cusp tips despite the high relief.

According to the original mesowear method (Fortelius and Solounias, 2000), cusp sharpness reflects abrasion and thus Bovini at Kanapoi likely included graze in their diet despite their low average mesowear score.

These results are supported by the isotopic analyses: at Kanapoi, one Bovini

85

individual was sampled and a δC13 value of -9.4‰ was obtained (Figure 10A and 10B).

At Allia Bay, the isotope signal is that of a C3-C4 mixed feeder: -5.9‰. The tooth enamel oxygen value associated with the Allia Bay Bovini individual sampled is -1.3‰, which is the lowest value within the Allia Bay bovids (Figure 10C). This is indicates that the individual was likely an obligate drinker. Although sample sizes are small in both cases,

Bovini clearly had different diets at Kanapoi and Allia Bay. These results are the opposite of our predictions of more open settings at Kanapoi than at Allia Bay.

Hippotragini

Living Hippotragini comprise both grazing and browsing species (Kingdon et al. 2013a).

Isotope data are not available for Hippotragini for any of the assemblages we analyzed.

No specimens from Allia Bay were suitable for the mesowear analysis. At Kanapoi, this taxon has a mixed diet, as shown by their mesowear score (2.25 ± 1.5).

Reduncini

The members of the Reduncini tribe had a graze-dominated, but mixed diet at Allia Bay.

The mean mesowear score for the tribe is 2.3 ± 2.1 (Figure 9). The only individual analyzed isotopically has a diet that can be interpreted in a similar fashion: -2.37‰

(Figure 10A and 10B). Its hypsodonty index is 1.10, the lowest amongst the bovid tribes from Allia Bay, which means they are classified as brachydonts. Its very high18O score of 4.69‰ (Figure 10C) is indicative of arid-adapted taxa, which likely use leaf-water as their main source of water (Levin et al. 2006). This contrasts with modern reduncins,

86

which are characterized by a high water dependency and select habitats close to standing water (Shipman and Harris 1988; Vrba 2006). In addition to error associated with using a single datapoint, multiple alternative explanations to this phenomenon exist. For example, the individual could be getting its water from shallow water, which is a lot more enriched in18O because of the higher impact of evaporation (Krishnamurthy and

Bhattacharya 1991). Modern reduncins are also grazers, therefore their diet may influence differently their their enamel18O values (Cerling et al. 2003).

87

Figure 10A

88

Figure 10B

89

Figure 10C

90

Neotragini

We scored a single Neotragini individual for mesowear. Its score is 2, which is within the range of the Neotragini from Kanapoi (1.57 ± 1.4, Figure 9). The hypsodonty indices are almost identical at Allia Bay and Kanapoi (1.49 and 1.51 respectively), at the border between brachydonty and mesodonty. Isotopic data could not be obtained for any of the assemblages.

Discussion

When the data from taxonomy, mesowear, ecomorphology and stable isotopes are combined, the signal from both the Allia Bay and the Kanapoi assemblages suggests highly heterogeneous environments, where open, intermediate and closed settings are represented in almost identical proportions. Open environments are overrepresented at

Kanapoi (33.31% vs. 25.81% at Allia Bay) and closed environments are marginally more common at Allia Bay (33.72% vs. 29.42%). Mursi environments are reconstructed as dominated by closed settings, but the results were often hindered by small sample sizes.

When methods are examined individually, the results vary, which highlights the importance of well-resolved taxonomic identifications investigated via multiple proxies.

Results also vary according to the taxon examined. Despite some differences mainly due to body size and differential preservation (i.e. higher frequencies of

Antilopini at Kanapoi), fossil bovid tribe abundances at Allia Bay and Kanapoi are statistically similar. Differential preservation is due to a wide range of factors including the properties of the bones themselves (bone size, bone density, etc.) and taphonomic factors (transport, weather, carnivore damage, sediment compression, etc.) (Lyman 1994;

91

Lyman et al. 1992). Feibel (2011) proposes that sedimentation rate and erosion were important factors in the preservation of fossil remains at Allia Bay: high sedimentation rates originally favored the preservation of organisms, but lower accumulation rates coupled with erosion negatively impacted the preservation of the fossils. Based on our own observations, hundreds of small mammal fossil remains were recovered at Kanapoi, but no more than a dozen small mammal remains were recovered at Allia Bay despite thorough excavation and screening (Hagemann 2010; Manthi 2006). Our taxonomic analyses show that a larger proportion of small-bodied taxa are preserved at Kanapoi than at Allia Bay. These small bovids are likely to inhabit forested setting (Kappelman 1988), which have acidic, organic soils that do not favor fossilization of skeletal remains. What we observe here echoes the Osteological Paradox, which exposes for bioarcheologists the problem of studying ancient health using only skeletons with traces of certain diseases.

The presence of small sized bovids at Kanapoi is a Taphonomic Paradox of sorts, as it is likely due to the fact that they preserve better in open environments rather than an indication of the presence of closed settings. We suggest that a thorough taphonomic study of the site may shed light on this issue.

In addition, astragalar ecomorphology results support the presence of many types of environments and Kanapoi and Allia Bay. One notable difference between the two assemblages is that postcranial ecomorphology results reveal that the proportion of

‘open‘ settings is higher at Kanapoi than at Allia Bay. The proportion of ‘forest’ habitats is also higher at Kanapoi. We suggest that this result can be further attributed to differential preservation of fossil remains between Allia Bay and Kanapoi. We propose

92

that methods applied on teeth, such as enamel stable isotopes and mesowear will be less affected by this bias because teeth preserve much better than bones in all environments.

All bovids were classified as either mesodonts or brachydont by the hypsodonty index, suggesting that the animals didn’t consume a lot of graze and/or lived in relatively mixed habitats. These results likely reflect the mismatch between evolutionary adaptation and the changing landscapes. There is variation in the results of the mesowear analyses, but overall, the mesowear scores for each tribe are similar at Allia Bay and Kanapoi.

When the Mursi specimens could be sampled, they showed more browse-dominated adaptations than in the two other assemblages.

One notable exception to this pattern is the tribe Aepycerotini. Mesowear analyses show that aepycerotins were grazing more frequently at Kanapoi than at Allia Bay. This result is consistent in the carbon and oxygen isotope analyses. The agreement between these two methods operating on different scales increases the robustness of the results.

Similarly, stable isotope values also showed a difference between Allia Bay and

Kanapoi Antilopini, which incorporated more grass into their diet at Kanapoi than at

Allia Bay. The reconstruction of the paleodiet of Antilopini only relies on stable isotopes since no individuals from the Allia Bay collection were suitable for analysis.

Bovini constitute another exception but show the opposite pattern. They appear to have been browsers at Kanapoi and mixed feeders at Allia Bay. These results highlight the heterogeneity of the environments and/or the dietary flexibility of Bovini during the

Pliocene in eastern Africa. However, in both cases, these results are hindered by small samples sizes thus both the mesowear and the stable isotope results may be misleading.

In addition, the mesowear scoring system used partially masks the effect of cusp

93

sharpness by summarizing both mesowear variables (cusp sharpness and occlusal relief) in a single score.

Other than these exceptions, results show overlap between bovid tribes at Allia

Bay and Kanapoi (e.g. Tragelaphini, Neotragini), both of which show consistently less depleted values than their counterparts from the Mursi Formation. The absence of grazers at Mursi is consistent with previous paleoecological reconstructions (Drapeau et al.

2014). Similarly, all taxa in the Mursi collection can be classified as being obligate drinkers, which is not the case at Allia Bay or Kanapoi, where several taxa are arid-adapted.

Overall, our paleoecological reconstructions partially align with what is currently known about Australopithecus anamensis behavior and morphology, but also provides additional insights into their paleoecology. In particular, using methods like mesowear and stable isotope on three faunal assemblages allows for a more nuanced reconstruction of the environments since all bovid tribes do not follow the same pattern. Despite being the earliest habitual bipedal hominin known, Australopithecus anamensis maintained more adaptations to an arboreal lifestyle than most later hominins, (Green et al. 2007;

Ward et al. 2013) and fed on C3 resources (Sponheimer et al. 2013). With the Savannah

Hypothesis, Dart proposed bipedal locomotion to be tied to the spreading of savannahs

(Dart 1925). More recent hypotheses are more nuanced, but still link the development of the first evidence of occasional bipedality in hominids like Ardipithecus (White et al.

2009b) to the progressive expansion of open habitats in eastern Africa around 8-5 Ma

(Lee-Thorp and Sponheimer 2015). At Kanapoi and Allia Bay, Australopithecus anamnesis would have found relatively open environments that also incorporate more

94

closed settings, which were well adapted to their diet and their still partially arboreal lifestyle.

Conclusion

We used Bovidae remains from the tribes Tragelaphini, Aepycerotini, Alcelaphini,

Antilopini, Neotragini, Bovini, Reduncini and Hippotragini to reconstruct of the paleoenvironments at Allia Bay as relatively grassy woodlands. Our analyses suggest that bovid remains from Allia Bay reconstruct the environments as broadly similar to the well-studied site of Kanapoi, where the environment is relatively open and heterogeneous

(Manthi et al. 2018; Ungar et al. 2018; Wynn 2000). Notable exceptions include the tribes Aepycerotini and Bovini, which could reflect the dietary flexibility of the tribe, heterogeneity of the environments, the limits of the method(s) used and/or the presence of different species of each tribe at Allia Bay and Kanapoi. Dental enamel stable isotope values for Aepycerotini and results from the ecomorphology analysis indicate that environments were more open at Kanapoi, where the most hominin remains were found, than at Allia Bay but further study is needed to confirm this interpretation.

The gradient from more open environments in the north of the Omo-Turkana

Basin to more closed conditions with Mursi in the south is consistent with conclusions derived from taxonomic analyses of bovids made by Bobe and colleagues for the period between 3.4 and 2.0 Ma (Bobe et al. 2007b). The idea of a gradient between the Omo

Valley, East Turkana and West Turkana - represented here by Mursi, Allia Bay and

Kanapoi - is also supported by the results of pedogenic carbonate analyses (Levin et al.

2011).

95

New data on the Allia Bay bovids has enabled the first faunal based comparison between A. anamensis sites in the Omo-Turkana Basin. This partial analysis of the fauna at Allia Bay suggests that the paleoenvironments were more open than previously thought. The implication is that A. anamensis is now resolutely associated with grassy woodlands in the Omo-Turkana Basin. Other studies have shown that the same hominin also lived in mosaic woodlands in the Awash Basin (Curran and Haile-Selassie 2016;

Kullmer et al. 2008; White et al. 2006), which perhaps represents the more wooded end of the spectrum of habitats selected by this species. Based on our results, we propose that the environments at Mursi were too closed for A. anamensis to inhabit. Further analyses of the complete fauna from the Allia Bay collection will provide a test of this hypothesis as well as an answer to our remaining interrogations on the paleoecology of this important hominin species.

Acknowledgements

We thank the Leakey Foundation, Sigma Xi Grants-in-Aid of Research, Explorers Club

Washington Group Inc., Evolving Earth Foundation, Cosmos Club Foundation and the

Lewis N. Cotlow Fund (to LD) for funding this research. We are grateful to the staff of the Ethiopian National Museum in Addis Ababa and the Nairobi National Museum. We thank the teams who recovered and curated the fossils that comprise the Allia Bay,

Kanapoi and Mursi collections.

96

Chapter 4: Pliocene suids (Mammalia, Cetartiodactyla) from Allia Bay and the paleoenvironments of Australopithecus anamensis in the Omo-Turkana

Basin

In preparation for submission in Fall 2018

Co-authors: René Bobe, Jonathan G. Wynn, W. Andrew Barr (Final list TBD)

Co-author contributions: LD conceived of this study. LD drafted the manuscript with feedback from all authors throughout the process.

Keywords: Suidae; paleoecology; Eastern Africa; taphonomy; stable isotopes;

Australopithecus anamensis

97

Abstract

A warm and wet period has been interpreted for eastern Africa around 4 million years ago resulting in widespread large, deep lake systems, such as the paleolake Lonyumun that filled the Omo-Turkana Basin. Relating such regional phenomena to more local changes and their impact on hominins continues to be one of the most important challenges in paleoanthropology. This paper contributes new data on Suidae from the early Pliocene site of Allia Bay, Koobi Fora Formation, Kenya and their implication for the reconstruction of the paleoenvironments of the earliest species of the genus

Australopithecus, Australopithecus anamensis, specifically addressing their association with humid settings. We compare fossils from the Allia Bay collections to assemblages from two penecontemporaneous localities, Kanapoi (4.17-4.07 Ma) and Mursi (4.2-c.3.9

Ma). We compare relative taxonomic abundances in the three assemblages, carbon and oxygen isotopes of mammalian tooth enamel as well as bone surface modifications

(taphonomy). Our results show that the vast majority of suid remains from Allia Bay belong to the subfamily Tetraconodontinae and that fossils attributed to Nyanzachoerus, and specifically to Nyanzachoerus kanamensis clearly dominate the assemblage. Grazers

Notochoerus are less common in the Mursi assemblage compared to the Allia Bay and

Kanapoi assemblages. Carbon isotopic data shows overlap between suids from the fossil

collections from Allia Bay and Kanapoi, predominantly mixed C3-C4 feeders. Mursi suids had a C3-dominated diet. Oxygen isotopic data shows obligate drinking behavior in Mursi suids and less so in suids from the two other collections. The analysis of bone surface modifications, particularly the weathering and polishing patterns, further suggests the

98

presence of a humid environment at Mursi. Overall, paleoenvironments from Mursi as reconstructed in this study displayed humid settings in which suids would have been able to thrive on varied diets and be able to stay in proximity to water. In addition, results reinforce the paleoecological similarities between Allia Bay and Kanapoi, both reconstructed as less humid. Allia Bay and Kanapoi are two only sites where A. anamensis was found in the Omo-Turkana Basin, suggesting an association with drier environments at the onset of the hominin lineage.

Introduction

The specific habitats preferences of the earliest species of the genus Australopithecus,

Australopithecus anamensis, are poorly understood relative to the species’ importance in our understanding of the human lineage and the emergence of obligate bipedalism. The species lived in eastern Africa about 4 million years ago (Ma), a period which has been interpreted as relatively warm, wet and associated with the formation of large lakes, such as the paleolake Lonyumun, which filled the Omo-Turkana Basin (Brierley et al. 2009;

Feibel 2011; Ring et al. 2018). This paleolake extended well beyond the edges of modern-day lake Turkana and had major influences on the climate and other environmental factors such as moisture levels and vegetation diversity (Trauth et al.

2010). This period also corresponds to the spreading of tropical C4 grasses in eastern

Africa (Cerling et al. 1998), a phenomenon attributed to the complex interaction of rifting processes (Sepulchre et al. 2006), the cooling of the Indian Ocean (Cane and Molnar

2001) and the intensification of high-latitude glacial cycles (deMenocal 2004). Cerling

99

and colleagues (2011) used carbon isotopic compositions of paleosols to demonstrate that environments were more closed in the Omo-Turkana Basin around 4 Ma than during the

Late Miocene or the Pleistocene. The consensus in the literature is that the Omo-Turkana

Basin was generally humid c.4 Ma, but that this varied at the local level. Reconciling climatic information collected at different scales (e.g., basinal to local) is still a major challenge in evolutionary biology (Marean et al. 2015; Vrba 2015).

This paper examines humidity within the basin at a local level by focusing on evidence from suid remains and bone surface modification from three important fossil collections c.4 Ma – the hominin sites of Allia Bay and Kanapoi, as well as the paleontological locality Mursi.

Background

Geological context

Australopithecus anamensis fossils were recovered from the Lonyumun Member of the

Koobi Fora Formation in the Allia Bay region, locality 261-1 (3°38' N and 36°16' E), hereafter referred to as “Allia Bay” (Figure 1). Although Allia Bay was recognized as a fossil-bearing locality during early reconnaissance of the Koobi Fora Formation, Feibel

(1988) was the first to formally identify and name the locality. He describes the sediments as sandstones containing highly polished fossilized bones and teeth. The Allia

Bay fossils were deposited in a fluvial context and represent a single depositional event

(Coffing et al. 1994). They were found in a bone bed situated below the Moiti Tuff, itself dated to 3.97+/- 0.03 Ma (McDougall & Brown, 2008) (Figure 2). The bone bed is about

100

200 m2 in extent and 20 cm thick and is estimated to be ~ 3.98 Ma(Coffing et al. 1994;

McDougall and Brown 2008).

Suidae fossils in paleoanthropology

Multiple lines of evidence have shown the association of the members of the suid genera

Nyanzachoerus and Notochoerus – at the exception of Notochoerus scotti - with the presence of humid environments during the eastern African mid-Pliocene (Bishop 1999;

Kullmer et al. 2008). Although studying the abundance of semi-aquatic or aquatic taxa such as Hippopotamidae and Crocodylidae would in theory be a more accurate indicator of these conditions, both of these taxa would provide a biased signal in practice using the data studied here because of differential preservation and different sampling strategies between the teams that worked at Allia Bay, Kanapoi and Mursi. For example, some teams did not routinely collect crocodilian fossils and bones of very large animals such as hippos were typically left in place. Additionally, Crocodylidae are homodont and shed their teeth, and it is therefore even more challenging to estimate their true abundance at a given site. Dental remains, particularly third molars, are one of the main indicators of taxonomic affinities in suids; therefore suid species can be distinguished even with highly fragmented teeth (Kullmer 1999), as is the case with the Allia Bay and Mursi assemblages. Suid remains are not typically affected by as many biases as other taxa and in the Koobi Fora monograph, J.M. Harris (1983b, p.215) mentions that: “[Suids were] sought intensively during the 1975 field season in connection with the preparation of a monograph on African Plio-Pleistocene suids and may therefore be slightly overrepresented in the collections.” Although this refers to a specific field season, it may

101

also possibly apply more broadly. Suidae constitute invariably one of the most abundant large mammal taxon in eastern Africa regardless of the local settings, often third in abundance to antelopes and zebras and representing a mode of about 5% of these living macromammal communities (Potts 1988). They are even more abundant in

Plio-Pleistocene assemblages from eastern Africa, often second in rank to bovids. Suids are the most abundant mammalian taxonomic family at Mursi (c.30% of mammal remains), and the second most common at both Kanapoi and Allia Bay (each c.20% of mammal remains).

The sensitivity of suids to climate change and the anagenetic trends characterizing the lineage (e.g., the elongation of the third molar) make them one the most useful taxa for the interpretation of biostratigraphy at hominin sites and establish correlations between geological units, because of their high abundance (Bishop 1994; Cooke and

Maglio 1972; Harris 1983a; Harris and White 1979; Reda et al. 2017; White 1995).

Researchers have pointed to parallels between suids and hominins in terms of

Plio-Pleistocene taxonomic diversification, dentition, diet, body size and overall generalist lifestyle (Bishop 1999; Hatley and Kappelman 1980; White 1995).

Three factors make reconstructing paleoenvironments based on fossil suids particularly challenging. First, in contrast to bovids, extinct and extant suid taxa do not tend to be analogous in terms of dietary and habitat selectivity. Harris and White (1979) explain that “[m]any works have failed to relate fossil material to the models of variation provided by extant African suid species”. Second, suids are flexible in their habitat and diet choices, it is thus important to consider a large sample size, like we are doing here.

And finally, suid taxonomy is complex, with conflicting taxonomic attributions made by

102

different researchers in the 1970s, most notably Harris, White and Cooke (Bishop 1999;

Boisserie et al. 2014).

Materials

The Allia Bay fossil assemblage contains 295 suid specimens, which are mostly isolated teeth. The fossils are housed at the Nairobi National Museum, in Kenya. Of these, we were able to identify 97 fossil remains at the specific level and 143 at the generic level.

No postcranial elements were found in association with craniodental material. Dental metrics follow Harris and White (1979).

Abbreviations

AB, Allia Bay; ER, East Rudolf (= East Turkana); KP, Kanapoi; KNM, Kenya National

Museum; Lt., left; MR or MUR, Mursi; Rt., right; U., unsided; WT, West Turkana.

Systematic Paleontology

CETARTIODACTYLA Montgelard, Carzeflis, and Douzery, 1997

SUIDAE Gray, 1821

Suids are omnivorous Cetartiodactyls that belong to the Superfamily Suoidea (Harris and

Liu 2007). Extant suids comprise seven genera and 12 species (Harris and Liu 2007), of

103

which three genera - Hylochoerus, Phacochoerus, Potamochoerus - and four species live in eastern Africa (IUCN 2017). Two subfamilies and three genera are represented in the

Allia Bay fossil assemblage: Nyanzachoerus, Notochoerus and Kolpochoerus.

TETRACONODONTINAE Lydekker, 1876

Taxa from the subfamily Tetraconodontinae, which are hypothesized to originate from

Eurasian Late Miocene suids, are not found in eastern Africa after the end of the Pliocene

(Cooke and Wilkinson 1978; Pickford 1986). The subfamily is defined by a reduction of the first and second premolars and an enlargement of the third and fourth premolars

(Harris and Liu 2007), although researchers point out that these traits are neither synapomorphies, nor do they apply to all species within the clade (Geraads and Bobe

2018b; Pickford 2014). Tetraconodontinae molar cusps are columnar and star-shaped to

H-shaped in cross-section, which are amplified expressions of the “three furrow” or

“furchen” pattern characterizing the cheek teeth of all suids (Hunermann 1968; Pickford

1986). Most of the suid remains from Allia Bay belong to this subfamily.

NYANZACHOERUS Leakey, 1958

Fossils attributed to Nyanzachoerus, and specifically Ny. kanamensis, dominate the Allia

Bay assemblage. Nyanzachoeres have sexually dimorphic canines of moderate size and large premolars, but relatively simple and bunodont third molars (Harris and White

1979). When worn, the dentine lakes form a stelatte pattern. Species from this genus

104

lived from c. 7.5 Ma to 2.85 Ma (Bishop 2010; Boisserie et al. 2014). At least six species are known: Ny. syrticus (= tutolus), Ny. devauxi, Ny. australis, Ny. kanamensis, Ny. khinzir and Ny. waylandi (= Ny. kuseralensis, following Boisserie et al. 2014) (Boisserie et al. 2014; Haile-Selassie et al. 2009; Harris and White 1979; Leakey 1958; Pickford

1989).

Kullmer (1999) argues that nyanzachoeres had a mixed diet based on their dental morphology, but stable isotope studies have demonstrated that some individuals can also

adopt either a C4 (Harris and Cerling 2002; Levin 2008) or a C3 diet (Drapeau et al.

2014). Nyanzacheorus kanamensis has been associated with “intermediate” environments, based on their cranial and postcranial remains (Bishop 1999). In addition,

Geraads and Bobe (2018b) suggest that Ny. kanamensis lived in an open environment based on its cranial features but the cranial remains associated to this species in the Allia

Bay collection are too fragmentary to assess this.

NYANZACHOERUS SYRTICUS Leonardi, 1954

Allia Bay material: 30341, Lt. hemi-mandible with teeth.

Nyanzachoerus syrticus (=tutolus) is a suid species mainly known from late

Miocene sites. Its latest appearance date is currently fixed at 5.23 Ma in the Upper

Member of the Nawata Formation (Leakey et al. 1996). These early suids have particularly large third and fourth premolars, and their molars are bunodont and brachydont with a relatively simple structure (Harris 1983b). Their cheek teeth are also

105

characterized by the presence of a cingulum. Ny. syrticus is smaller than Ny. kanamensis

(Harris and White 1979).

KNM-ER 30341 (Figure 11A) is a complete weathered maxilla with worn teeth.

On the maxilla, the third molar is reduced compared to the M2 and the premolars, which led us to identify it as cf. Ny. syrticus. This specimen could either be a late-persisting Ny. syrticus or a primitive-looking Ny. kanamensis.

NYANZACHOERUS KANAMENSIS Leakey, 1958

3 Allia Bay material: 6166, Lt. M fragment; 6169, Rt. M3 fragment; 6170, Rt. M3 fragment; 6180, U. M3 fragment; 17879, Lt. M3 fragment; 19541, Lt. M3 fragment;

30367, complete Rt. P3; 30391, associated teeth including an M3; 42642, complete Rt.

M3; 42636, Lt. M3 fragment; 42644, Rt. M3 fragment; 42848, Rt. M3; 42849, complete

3 Rt. M2-3; 42857, complete Lt. M2; 42870, Rt. M2 fragment; 42871, complete Lt. M ;

3 3 42874, complete Lt. M ; 42910, Rt. M3; 42913, complete Lt. M ; 42914, Rt. maxillary

3 fragment with M ; 42915, Lt. M3 fragment; 42916, complete Lt. M3; 42929, complete

Rt. M3; 42937, U. M3 fragment; 42941, complete Rt. Hemi-mandible with teeth; 43024,

1 2 complete Lt. M ; 43025, complete Rt. M ; 43029, Lt. M1 fragment; 43033, Lt. M1 fragment; 43034, complete Lt. M1; 43041, complete Rt. M2; 43047, cranial fragment;

3 3 43250, Rt. M ; 43259, complete Lt. M2; 43270, complete Lt. M ; 43480, complete Lt.

106

Figure 11: (A) cf. Nyanzachoerus syrticus, KNM-ER 30341, one of two very weathered left maxillary fragments with P3 to M2 (M3 present in another fragment, not illustrated), superior view. (B) Nyanzachoerus kanamensis, KNM-ER 43855, left 3 mandibular fragment with M3 in place. (C) Notochoerus jaegeri, KNM-ER 42643, right M . (D) Kolpochoerus heseloni,

KNM-ER 42641, left M3 (E) Notochoerus euilus, KNM-ER 30350, left mandibular fragments with M2 and M3. (F) Nyanzachoerus kanamensis, KNM-ER 43564, almost complete right hemi-mandible with P4-M3. Scale bar= 2 cm for (A)-(E). A second scale bar = 3 cm for (F).

107

3 M1; 43544, complete Rt. M ; 43564, almost complete Rt. hemi-mandible with teeth ;

43565, complete Lt. hemi-maxilla with teeth; 43566, complete Lt. hemi-mandible with teeth; 43855, complete Lt. hemi-mandible with teeth; 43856, complete Lt. hemi-mandible with teeth; NA (LD-18/ER-88), M3 fragment.

Nyanzachoerus kanamensis is a relatively large suid with large and bunodont posterior premolars. Their molars lack a cingulum, have column-like cusps joined by small central pillars and have extended talons and talonids compared to the equivalent teeth in Ny. syrticus. When worn, the dentine lakes of the molars take a relatively simple star shape. Ny. kanamensis have highly sexually dimorphic crania (Harris and White

1979). Some researchers use Ny. pattersoni to refer to specimens from Kanapoi that overlap in morphology with Ny. kanamensis (Harris et al. 2003), while other researchers consider Ny. pattersoni a junior synonym of Ny. kanamensis. The latter is the position we adopt here, following Geraads and colleagues (2013). Fesseha also argues that the species should instead be classified within the Notochoerus genus (Fesseha 1999).

With 70 craniodental fossil remains attributed to species, Ny. kanamensis is the most common suid at Allia Bay as well as at Kanapoi (Geraads and Bobe 2018b).

Noteworthy specimens include KNM-ER 43564, an almost complete right hemi-mandible without its ascending ramus (Figure 11F). The M3 is a recently erupted tooth. The M2, M1 and P4 are present, but the P3 and P2 alveoli are empty. The mandibular symphysis is intact and is not as deep as in Notochoerus. The teeth are relatively brachydont. The 3rd molar talonid is not as elaborate as in No. jaegeri.

108

NOTOCHOERUS Broom, 1925

Researchers noticing similarities in dental and mandibular remains in species of

Notochoerus and Nyanzachoerus have suggested that notochoeres are putative descendants of the nyanzachoeres (Bishop 2010; Cooke 1978; Harris and White 1979;

Kullmer 1999), although we know that late nyanzachoeres are synchronous with early species of notochoeres (Reda 2011). Differences with nyanzachoeres include overall body size, the shape of worn cusps and relative proportions of the cheek teeth.

Notochoerus are larger suids, but their third and forth premolars are reduced compared to their predecessors. Their molars became larger, especially the third molar, which is both more complex and more hypsodont than in Nyanzachoerus (Harris 1983b). The worn cusps are “H” shaped. Five species are currently recognized, No. jaegeri, No. euilus, No. capensis, No. clarki and No. scotti, but taxonomy is still debated. The genus would have lived in eastern Africa from about 4.35 Ma to 1.8 Ma (Bishop 2010; Reda 2011).

On the basis of their postcranial ecomorphological adaptations, No. euilus are associated with closed environments (Bishop 1999). Notocherus jaegeri and No. scotti have yet to be studied. In addition, most researchers argue that their very hyposont teeth, cranial morphology and large body size are more concordant with a highly abrasive, grass-dominated diet (Geraads and Bobe 2018b; Harris and White 1979).

109

NOTOCHOERUS JAEGERI Coppens, 1971

Allia Bay material: 17372, complete Rt. M2; 30277, associated tooth fragments; 30279,

3-4 3 U. P ; 42643, complete Rt. M ; 42856, complete Rt. M3; 43859, Lt. M3 fragment.

Notochoerus jaegeri (= plicatus) presents many similarities to Ny. kanamensis and the species was originally attributed to the genus Nyanzachoerus (Coppens 1971).

Some authors argue it should be reclassified as Nyanzachoerus (Reda 2011; Reda et al.

2017). This species shows reduced third and fourth premolars as well as longer and more complex third molars compared to nyanzachoeres (Harris and White 1979). The six specimens that could be attributed to this species are teeth with extended posterior projections and that were larger than in most Nyanzachoerus. Other specimens have affinities with both Nyanzachoerus and Notochoerus, which supports an anagenetic relationship between these genera. For example, KNM-ER 43849 is similar in general morphology to Ny. kanamensis, but is much larger and displays H-shaped wear in addition to being relatively bunodont.

NOTOCHOERUS EUILUS Hopwood, 1926

Allia Bay material: 30352, complete Lt. hemi-mandible with teeth; 34842, complete Lt.

P3; 42932, complete Lt. P3; 42981, complete Rt. P3; 42985, complete Lt. M2; 43012,

3 complete Rt. P ; 43474, complete Rt. M2; 43483, complete Rt. M1; NA (ER-3686), mandibular fragment with teeth.

110

Notochoerus euilus has reduced premolars and dumb-bell shaped cups as well as more complex and more hypsodont second and third molars than its predecessor No. jaegeri (Leakey 1958). However, some authors have argued that this does not warrant specific distinction with No. jaegeri (Harris 1983b; Kullmer et al. 2008).

Eleven cranial and dentals specimens in the Allia Bay assemblage are attributed to this species. For example, KNM-ER 30350 is a left mandibular fragment with the M2 and

M3 in place (Figure 11E). The teeth are large and the pillar arrangement is typical for this species. This specimen is not as hypsodont as No. scotti. The metaconid is slightly larger than the protoconid (Harris and White 1979) and there are two pillars in between each pair of cusps (i.e., double medial pillars), a trait sometimes seen in dental remains from this species. The specimen also exhibits three pairs of talonids, whereas the species description mentions two to four pairs. KNM-ER 30338 is a distal left maxillary fragment with the third molar in place. The tooth is more hypsodont than Nyanzachoerus (Table 4), talonid pillars are at almost the same height as the rest of the tooth, whereas they are lower in No. jaegeri. The tooth is also too wide to be attributed to No. scotti, which are characterized by their narrow teeth. We identify it as No. cf. euilus, but while it corresponds to the species description in terms of crown height, it is only two pillars wide, making it particularly narrow. No. euilus typically have wide M3s, with multiple buccolingually-placed pillars.

111

NOTOCHOERUS SCOTTI Leakey, 1943

Allia Bay material: 23292, complete Lt. M3.

Notochoerus scotti is the youngest and most derived species of this genus, with teeth markedly more hypsodont than in earlier species (Bishop 2010). Only specimen

KNM-ER 23292, a complete left third molar attributed to this genus in the Allia Bay assemblage, was sampled for isotopic analyses.

SUINAE Gray, 1821

Suinae fossils, which date back to the end of the middle Miocene of Europe, are thought to appear much later in Africa, at the beginning of the Pliocene (Harris and Liu 2007).

The subfamily shares an extension and increase in hypsodonty in the third molar (Harris and White 1979). Suines persists today and comprises all pigs including the three African wild pigs: the forest hogs (Hylochoerus), bush pigs (Potamochoerus) and warthogs

(Phacochoerus).

KOLPOCHOERUS Van Hoepen and Van Hoepen, 1932

This genus bears resemblances with the modern bushpig Potamochoerus. Both taxa are characterized by extended zygomatic arches (Bishop et al. 2006). The genus is further described as having a dorsally expanded neurocranium and an inflated mandibular body

(Souron et al. 2013). The genus, which lived in Africa from the early Pliocene (c. 5 Ma)

112

to the Middle Pleistocene (Bishop et al. 2006; Brunet and White 2001; Harris et al. 1988;

Souron et al. 2013), is most abundant in the Omo-Turkana Basin between 2.53 and 1.87

Ma (Rannikko et al. 2017). Currently, eight species of Kolpochoerus are recognized: K. deheinzelini, K. afarensis, K. cookei, K. millensis, K. heseloni (=K. limnetes, including K. paiceae and K. olduvaiensis), K. majus, K. phacochoeroides and K. phillipi (Brunet and

White 2001; Cooke and Wilkinson 1978; Geraads 2004; Geraads et al. 2004; Haile-

Selassie and Simpson 2013; Harris and White 1979; Hendey and Cooke 1985; Souron et al. 2013; White and Harris 1977).

Only three specimens in the Allia Bay collection are attributed to the genus

Kolpochoerus (or cf. Kolpochoerus). KNM-ER 30377 comprises a lower third molar as well as an incisor, a second premolar and various tooth fragments. The third molar is relatively brachydont, with a talonid that is less extended than in Ny. kanamensis.

KOLPOCHOERUS HESELONI Leakey, 1943

The species is described as showing expanded and hyposodont third molars and reduced premolars (Harris and White 1979). We confidently attribute a single specimen

(KNM-ER 42641) to the species Kolpochoerus heseloni (Cooke 1997). This complete left lower third molar (Figure 11D) is too massive and triangular to be attributed to Ny. kanamensis and the talonid is too complex to be K. afarensis (Cooke and Wilkinson

1978). The anterior cingulum is prominent, but less so than the major cusps of the tooth

(Harris and White 1979). This specimen is possibly intrusive.

113

This species is generally associated with a grazing diet (Bishop et al. 2006;

Cerling et al. 2015), but their mesodonty (Harris and White 1979) and the isotope value reported here suggest K. heseloni had a more mixed diet at its emergence c. 4 Ma. Bishop and colleagues observed that K. heseloni frequently co-occurs with hominins in eastern

Africa, but not in South Africa, a difference they hypothesize to be linked to ecology amongst other factors (Bishop et al. 2006).

Carbon and oxygen isotopes of mammalian tooth enamel

Enamel carbon isotopic data is commonly used as a dietary indicator in ecological and paleontological studies. This method is based on the fact that there exists three different pathways in terrestrial tropical plants photosynthesis. Warm-growing-season and low-elevation grasses and sedges use the C4 photosynthetic pathway for carbon fixation.

Those that use the C3 pathway include a much broader array of vegetation: trees, shrubs, bushes, sedges and cold-growing-season grasses. In tropical environments, plants that use

13 the C3 photosynthetic pathway yield values that are C-deplete compared to those that use the C4 photosynthetic pathway (respectively, C3 plants around -22 to –35 per mil (‰) compared to C4 plants at about -10 to -15‰) (Cerling et al. 2015; Koch 1998; Lee-Thorp et al. 1989). Crassulacean acid metabolism (CAM) is reserved to succulents, thus generally limited to desert environments (Lüttge 2004).

114

3 3 Genus Species Specimen M M M M3L M3W M3 Wear number L 3 Crown H Crown H (0-3) W Kolpochoerus

sp. 30377 - - - NA NA 2.3 1

heseloni 42641 - - - 5.7 2.8 2.2 2

Notochoerus

jaegeri 42643 6.2 3.4 0.9 - - - -

42856 - - - 8.8 3.2 0.9 3

43849 - - - 5.4 2.4 1.4 2

euilus 30350 - - - 7.6 1.9 3.2 2

cf. euilus 30338 - - - NA NA 3 0

scotti 23292 - - - 6.6 2.1 3.5 2

sp. 42645 - - - 3.1 2.2 0.7 3

Nyanzachoerus

syrticus 30341 - - - 5.1 2.8 2.3 3

cf. syrticus N

30275 A NA 1.3 - - - -

kanamensis 6169 - - - 3.6 2.6 1.8 0

6170 - - - 4.1 2.4 2.2 1

19541 5.1 3.1 2.1 - - - 1

30391 - - - NA NA 2.4 0

42636 - - - 3.5 2.3 1.2 3

42642 - - - 3.7 2.45 1.7 2

42644 - - - NA NA 1.2 2

42848 - - - NA NA 1.4 2

42849 - - - 4.9 2.2 2.2 2

42871 4.6 2.85 1.7 - - - -

42874 4.4 2.9 1.9 - - - -

42910 - - - 4.4 2.1 1.9 1

42913 5.1 3.2 2.3 - - - -

42915 - - - 4.4 2.3 1.4 2

115

42916 - - - 5.1 2.9 1.1 3

42929 - - - 5.3 2.3 1.8 1

42941 - - - 5.7 2.6 2.1 1

43250 5.8 2.6 1.1 - - - -

43270 6.6 4.2 1.2 - - - -

43544 5 3.1 1 - - - -

43855 - - - 5.3 2.5 1.4 2

43856 - - - 5.4 2.7* 1.8 2

Measurements: M#L molar number length; M#W molar number width; M#Crown H molar number crown height; * value is an estimate.

Table 4: Measurements (in mm) for all Suidae molars identified to at least the genus level in the Allia Bay assemblage

Carbon isotope-based paleodietary studies are concerned with the types of vegetation (i.e., C3 or C4) ingested by herbivores and omnivores. The enamel bioapatite

13 in the teeth of animals reflects the carbon isotopic composition of their diet. δ C values

preserved in enamel that are -1‰ or higher are found in animals with pure C4 diets, values of -8‰ or lower characterize C3 feeders, and values in between these two extremes indicate a mixed C3-C4 diet (Cerling et al. 2015).

The oxygen isotopic composition of mammalian enamel bioapatite is another indicator of paleoecology, but environmental interpretations are complex, with a number of biotic and abiotic factors involved during the incorporation of oxygen environmental

116

water into the enamel structure (Luz et al. 1984). Factors influencing the composition of oxygen in water include temperature, altitude and distance from the oceanic source region, i.e. “continental effect” (Krishnamurthy and Bhattacharya 1991). Within animals oxygen isotope ratios reflect body water, which is itself linked to physiology, drinking behavior and leaf water (Koch 1998; Sponheimer and Lee-Thorp 1999). In obligate drinkers, which are animals that drink water directly from a meteoric source water, the tooth enamel reflects the oxygen isotopic composition of the water source which may reflect that of rainfall contributing to that source, modified to some degree by evaporation

(Dansgaard 1964). Non-obligate drinkers, or arid-adapted taxa, can survive without direct access to a body of water and instead obtain water from the plants (leaves, fruits, or other parts) they ingest (Ayliffe and Chivas 1990), which are 10-30‰ enriched compared to the meteoric source water (Yakir 1998). The oxygen isotopic composition of their enamel will thus be enriched compared to that of rainfall and influenced by the type and part of the plant they ingest. Since the isotopic composition of local source water is influenced by rainfall and other local climatic factors, geographic locations or periods are not directly comparable (Koch 1998).

Bone surface modifications

A number of paleoecological methods commonly used to analyse bovid remains cannot be easily translated for use on suid remains. One such method is the analysis of mesowear

(Fortelius and Solounias 2000), for which a method specific to suid teeth has yet to be developed. The paucity of postcranial remains in our samples prevents us from using

117

ecomorphological approaches (But see Bishop 1994; Bishop et al. 2006). To avoid over-reliance on a single line of evidence, we used surface modifications on post-cranial remains as a complementary indicator of local humidity, which is particularly relevant in the absence of more direct paleoenvironmental indicators, such as pollen data (Bonnefille et al. 2004).

Taphonomy is the discipline that studies the many ways in which processes can affect an organic remain peri-mortem or post-mortem (Efremov 1940). Taphonomists study the different traces on post-cranial remains or other substrates and attempt to tie them to a causal agent (Gifford 1991). Recognizing traces traditionally demands expert knowledge gained through experience, but recent studies have also shown that automating this process is now possible (Pante et al. 2017).

Here, we focus on agents and traces that can be more directly understood in terms of ecology. As such, we follow the definition proposed by Behrensmeyer and Kidwell

(1985), which emphasizes the role of taphonomy in paleoecology: “the study of processes of preservation and how they affect information in the fossil record”. We describe each agent briefly below but suggest referring to the Atlas of Taphonomic Identifications for further details, including many pictures (Fernandez-Jalvo and Andrews 2016).

Abrasion (or polishing) is caused by mobile particles in air or water and causes the rounding of an osteological element, and in most extreme cases the surface becomes smooth and glossy (Brain 1967; Lyman 1994; Vorrhies 1969). Actualistic studies have shown that most abrasion happens pre-fossilisation. Post-fossilisation abrasion can also be distinguished from pre-fossilisation abrasion when it occurs (Andrews 1995;

Thompson et al. 2011). A high abrasion level in an assemblage corresponds to a long

118

time or distance dragged in water (Behrensmeyer 1982). Many scoring systems have been suggested (Cáceres et al. 2012; Coombs and Coombs Jr 1997; Hunt 1978).

Meat-eating animals also leave traces on bones when they chew on them. Other animals such as rodents and large herbivores are known to occasionally leave traces on bones, but we did not identify any of these traces in the Allia Bay, Kanapoi or Mursi assemblages. Insects, termites in particular also leave traces on bones (Backwell and d'Errico 2001; Backwell et al. 2012; d'Errico et al. 2001), but experts conclude that the behavior is linked to a nitrogen deficiency and cannot be directly linked to characteristics of the habitat (Fejfar and Kaiser 2005). Here, we focus on traces made by mammalian carnivores (Marean and Spencer 1991; Montalvo et al. 2007) and crocodiles (Njau and

Blumenschine 2006). Tooth marks on bones can be used in combination with the presence of bones or teeth to assess the likely presence of carnivores at a site. Enzymes and gastric acids in the prey’s digestive tracts can also chemically alter the bones they ingest (Kitching 1963; Shipman 1981; Sutcliffe 1970). Bones affected by digestion are perforated and reduced in size (Kolska Horwitz 1990).

Linear traces occur when a bone lying on the surface and is trampled (Andrews and Fernández-Jalvo 2012; Blasco et al. 2008). Trampling is widely studied because of its resemblance to cutmarks (Reitz and Wing 1999). Experts have noted that traces caused by trampling are more likely to be U-shaped in cross-section than cutmarks

(Cáceres et al. 2012), shallower (Fiorillo 1989), and are often parallel to each other and found on the diaphysis rather than at the end of bones (Andrews and Fernández-Jalvo

2012). A high trampling level indicates that bones layed on the surface for a long time

119

and thus were buried slowly (Blasco et al. 2008; Olsen and Shipman 1988; Reitz and

Wing 1999).

Weathering, or the effect of weather (rain, sun, wind) on fossilized remains was formally described by Behrensmeyer (1978). The natural role of weathering is to degrade bones in order to recycle their nutrients, and thus it separates organic and inorganic components of bones (Reitz and Wing 1999). A highly weathered bone will be cracked with scales of bone that peel off (Eberth et al. 2007; Tappen 1969). According to

Behrensmeyer (1982), the weathering level, evaluated from 0 to 5, can be used to estimate time before burial, but that this is highly dependent on the local weather conditions. She interprets a bone assemblage with different degrees of weathering as the sign of a slower, more progressive burial (Behrensmeyer 1978).

Root etching causes dendritic traces on bones during the growth and decay of roots (Behrensmeyer 1978; Lyman 1994; Phoca-Cosmetatou 2005). The exact ecological interpretation of this event is misunderstood, despite its high frequency in the fossil record. It is generally interpreted as being linked to vegetal richness of the soil as well as to time before burial (White 1992), but it is understudied and thus poorly understood

(Tjelldén et al. 2015).

Corrosion (or dissolution) is a chemical or biochemical process leading to bone loss and deformation. It takes place when the skeletal remain is buried under water or in acidic sediment (Auguste 1994; Behrensmeyer 1984). Corrosion is notoriously difficult to identify macroscopically (Denys 2002) due to its similarity to the effects of digestion, polishing or weathering (Andrews 1995).

120

Methods

Taxonomy (systematic paleontology)

We compared the proportions of taxa present within each fossil assemblage by analyzing the data statistically using a Monte-Carlo simulated Fisher’s Exact Test with 5,000 iterations with Bonferroni control. We used the habitat preferences of genera as presented in the scientific literature (Recently summarized in Rannikko et al. 2017). Analyses by species give almost equivalent results, since most of the Nyanzachoerus fossils belong to the species Ny. kanamensis and most of the Notochoerus fossils belong to No. jaegeri.

Carbon and oxygen isotopes of mammalian tooth enamel

We examined a total of 67 carbon and oxygen stable isotope values obtained from faunal dental enamel from Allia Bay, Mursi and Kanapoi, including new (n=31) and published data (n=36) (Blumenthal et al. 2017; Cerling et al. 2015; Drapeau et al. 2014;

Schoeninger et al. 2003). The genera analyzed are Nyanzachoerus (n=50), Notochoerus

(n=11) and Kolpochoerus (n=1). The samples proveniences are as follows: Allia Bay

(n=17, all new values), Kanapoi (n=10, including 2 new values) and Mursi (n= 40,

3 including 12 new values). Molars, particularly M and M3 were sampled preferentially to avoid juvenile dietary signal in teeth.

We used a high-speed rotary drill fitted with a small diamond bit to clean the surface of each tooth before collecting approximately 10 to 15 mg of enamel powder. The

121

drill was cleaned between each sample to avoid contamination. The enamel powders were imported to the United States following the guidelines of the National Museums of

Kenya and Ethiopia.

We treated the samples of enamel powder with 1 M acetic acid-calcium acetate buffer to remove any secondary carbonates and organic matter. We performed a series of rinses with deionized water (Koch 1998). Samples were dried at 25 °C for 24h. Samples were reacted with 103% phosphoric acid and analyzed in a ThermoFisher Scientific Delta

V Advantage Isotope Ratio Mass Spectrometer at the stable isotope laboratory at the

University of South Florida. Replicate analyses of these standards within a given run deviated by less than 0.1‰. We report the results using the standard per mil (‰) notation:

13 18 3 δ C or δ O = [Rsample/Rstandard - 1] x 10 ‰

13 12 18 16 where Rsample is the C/ C or O/ O ratios of the sample and Rstandard is the ratio of the reference standard. δ13C and δ18O values of tooth enamel are reported on the VPDB scale, which was normalised using replicate analyses of IAEA and internal reference materials NBS-18 (δ13C= -5.01‰ and δ18O=-23.01‰) and NewCar (δ13C= 4.04‰ and

δ18O=-3.38‰).

We used the Mann-Whitney test to explore differences between fossil assemblages as well as a Kruskal-Wallis test with Dunn test adjusted with the

Benjamini-Hochberg method.

122

Bone surface modifications

Abrasion, carnivore damage (tooth marks and digestion), trampling, weathering, corrosion and root etching were evaluated by direct observation or with a 20x or 30x handheld lens in artificial light. Each agent is recorded according to its own system.

When possible, we used stages of intensity defined in the literature. The intensity stages we used are as follows:

Abrasion: scoring system from 0 to 3 (Fiorillo 1988)

Carnivore tooth marks: amount and type of tooth mark (punctures, pit, score or

furrow. See Pobiner 2008 for definitions) is recorded for each specimen

Weathering: scoring system from 0 to 5 (Behrensmeyer 1978)

Root etching, crocodile tooth marks, insect traces, corrosion and digestion:

Presence/absence

We restricted our study to post-cranial remains >5 cm in length. This allowed us to reduce bias since some taphonomic agents are more difficult to observe on small fragments (e.g., carnivore-induced damage). We evaluated a total of 907 specimens: 295 from Allia Bay, 420 from Kanapoi and 192 from Mursi. We used a chi-square to show that there is no significant difference between the three collections in specimen size

(p=0.4). Similarly, there was no difference between groups in the extent to which fossil surface was obstructed by matrix (evaluated by presence/absence) (p=0.1246). We evaluated our results for each taphonomic agent using the chi-square (with p-value computed by Monte Carlo simulation) and performed Fisher’s exact tests with Bonferroni control to see which groups differed.

123

All statistical analyses were performed in R 3.4.3.

Results

Suid abundance data (Figure 12) by genus are similar in the Allia Bay and Kanapoi assemblages and the differences are not statistically significant (p= 0.05 with Monte

Carlo Fisher’s exact test with 5000 simulations). Within both assemblages, Notochoerus, a grazer, is relatively common (32.2% and 24.8% of identifiable suids within the Allia

Bay and Kanapoi collections, respectively). On the contrary, the vast majority of identified suid in the Mursi collection (96.5%) are Nyanzachoerus, a genus known to

have a more mixed C3-C4 diet. The genera from the Mursi collection significantly differ from those represented in the Allia Bay and Kanapoi collections (in both cases, p= 0.0002 with Monte Carlo Fisher’s exact test with 5000 simulations).

124

Figure 12: Comparison of the suid taxonomic composition between the fossil assemblages of Allia Bay, Kanapoi and Mursi

Suids δ13C values in general are statistically different between the samples from the Allia Bay, Kanapoi and Mursi assemblages (Kruskal-Wallis chi-squared=16.808, p=0.001), but the post-hoc Dunn test adjusted with the Benjamini-Hochberg method show significant differences between samples from Allia Bay and Mursi (p= 0.001) and

Kanapoi and Mursi (p=0.008), but not Allia Bay and Kanapoi (p=0.899). The ratios of C4 feeders to mix C3-C4 feeders to C3 feeders for the suids from the three assemblages are as follows: Mursi 1:29:10; Allia Bay 3:13:1; Kanapoi 1:9:0.

The sample of Nyanzachoerus is large enough to allow for robust statistical comparisons between the fossil collections from Allia Bay (n=9), Mursi (n=33) and

Kanapoi (n=5). The δ13C values of enamel from the early Pliocene Allia Bay and

Kanapoi assemblages are indicative of a diet that is, on average, evenly partitioned

between C3 and C4 plants, with Nyanzachoerus having near overlapping values at Allia

Bay (range = -5.9‰ to 1.5‰, mean= -3.37‰, sd=1.71, n=9, Figure 13A) and Kanapoi

125

(mean= -4.39‰, range = -6.5‰ to 1.94‰, sd=1.62, n=5, Figure 13A). The δ13C values of the fossils from the Mursi collection are generally lower than zero and indicate either a diet dominated by C4 plants, or a mixed C3-C4 diet (Figure 13A). Although the results from the Mursi suids are statistically significant from those from Allia Bay

(Mann-Whitney, p=0.002), they do not differ from those from the Kanapoi values

(Mann-Whitney, p=0.083). Our results suggest that Nyanzachoerus had overlapping diets at Allia Bay and Kanapoi (Mann-Whitney, p=0.423), but the specimens have slightly more enriched values at Kanapoi, indicative of a graze-dominated diet. One sample from an unidentified suid produced fell into the grazing category (>-1‰). The average δ13C value in Nyanzachoerus enamel is highest in the Mursi fossils (mean= -6.43‰, range= -11.5 to -1.7, sd=2.53, n=36, Figure 13A) and the range between the minimum and maximum δ13C values is also the largest in the Mursi samples.

The δ13C values for Notochoerus are similar in the Allia Bay (range = -6.3‰ to

0.3‰, mean= -2.38‰, sd=2.41, n=7) and Kanapoi (mean=-3.54‰, range= -5.47 to -1.25‰, sd=2.13, n=3) fossils sampled. Some of the most enriched values are

associated with the No. euilus and No. scotti from Allia Bay, indicating that C4 plants were important in the diets of these animals. Contrary to expectations, Notochoerus from

Kanapoi did not firmly fall into the C4 category. Their diet would be more accurately described as a mixed C3-C4 diet dominated by C4 plants. At Allia Bay, Notochoerus individuals had diets ranging from mixed to a graze-dominated. The δ13C results are not statistically different in the Allia Bay and Kanapoi notochoeres sampled (Mann-Whitney,

126

p=0.383).

The δ18O results follow similar trends. Both published and new δ18O values for

Notochoerus in the Allia Bay assemblage range from -2‰ to 2.3‰, with a mean of 0.1‰

(sd=1.71, n= 7) and for Nyanzachoerus, the range is -2.9‰ to 0.15‰ and the mean is -1.1 ‰ (sd=0.91, n=9, Figure 13B). The mean δ18O value for Notochoerus from the

Kanapoi collection is slightly lower with -1.31‰ (range= -2.19 to 0.1, sd=1.71, n=3), but the value for Nyanzachoerus is less depleted at -0.25‰ (range= -2‰ to 1.19‰, sd=1.21, n=5, Figure 13B). The δ18O values are the most negative in the Mursi assemblage, with an average of -2.46‰ for Nyanzachoerus (range= -6‰ to 0.1‰, sd=1.43, n=36, Figure

13B).

127

Figure 13A

128

Figure 13B

Figure 13: 13C and 18O values of Suidae enamel by fossil assemblage. Data from

Drapeau et al. (2014), Harris et al (2003), Cerling (2015), Schoeninger et al. (2003), and this study. (A) Box-and-whisker plots of 13C values (B) Box-and-whisker plots of

18O values.

129

The only Kolpochoerus specimen suitable for analysis is from Allia Bay. Its δ13 and δ18O values, -9.82‰ and 0.1‰ respectively, are completely different from the rest of the suid assemblage. This taxon is also rare in the assemblage thus it is possibly intrusive.

The genus Kolpochoerus had also just emerged c. 4 Ma, so it is possible that they were exploiting the more closed settings left aside by Notochoerus and Nyanzachoerus. This hypothesis should be tested if a larger sample can be obtained.

Unidentified suids from the Mursi assemblage follow the trend, except for the molar fragment that was also an outlier for δ13C results. The two unidentified suids from the Kanapoi sample fall within the grazing range. No unidentified suids from the Allia

Bay collection were analyzed using stable isotopes.

Fossil collections were no different in terms of insect damage (p=0.48), trampling

(p=0.21), digestion (p=0.66) or root etching (p=0.354). Visual signs of chemical corrosion were significantly different between the three collections (χ2=21.679, p<0.001,

Figure 14A). Results were particularly influenced by the higher than expected amount of specimens affected by corrosion at Mursi, as illustrated in Figure 14A. The post-hoc test revealed no significant difference in corrosion between Allia Bay and Kanapoi. Both assemblages were however significantly different from the Mursi fossils. The levels of polishing were also different between the collections (χ2=17.924, p=0.005, Figure 14B.).

Further testing revealed differences in polishing between Allia Bay and Kanapoi

(p=0.029) as well as between Mursi and Kanapoi (p=0.029), but not between Allia Bay and Mursi. Results were influenced by the fact that higher levels of abrasion are very well

130

represented in the Mursi assemblage. The chi-square showed significant differences between fossil collections in crocodile-induced damage (χ2=9.3091, p=0.031), but none of the pairwise differences were significant. Carnivore tooth marks were also significantly different between faunal assemblages (χ2=29.888, p=0.008, Figure 14C.).

Post-hoc testing showed no significant differences between Allia Bay and Mursi, but significant differences between the two other pairs (Allia Bay-Kanapoi, p=0.001;

Kanapoi-Mursi, p=0.001). Figure 14C shows that fossil remains at Kanapoi have more carnivore-induced tooth marks than expected. When carnivore damage is evaluated by presence/absence, the results are unchanged. The assemblages are different overall

(χ2=13.005, p=0.016). In addition, there is no difference between Allia Bay and Mursi, but there are differences between Allia Bay and Kanapoi (p=0.001) and Kanapoi and

Mursi (p=0.01). Weathering levels were also found to be significantly different between assemblages (χ2=35.475, p<0.001, Figure 14D). Mursi and Allia Bay were not found to be statistically different, but the two other pair-wise comparisons were (Allia

Bay-Kanapoi, p=0.006; Kanapoi-Mursi, p<0.001). Specimens without any traces resulting from weathering were less common than expected at Kanapoi and more common than expected at Mursi, as shown in Figure 14D.

131

132

Figure 14A.

Figure 14B.

133

Figure 14C.

134

Figure 14D.

Figure 14: Mosaic plots describing the bone surface modifications from selected tahonomic agents: A) Corrosion; B) Polishing; C) Carnivore Damage; D) Weathering. The hue indicates the residual’s sign (blue for positive and red for negative). The saturation (“colorfulness”) indicates the size of the residuals. A high saturation is used for large residuals and a low saturation for small residuals. The value (amount of grey or lightness) indicates significance. Dark colors indicate non- significant results and lighter colors indicate significant results.

135

Discussion

Our analyses focused on the difference between three fossil localities, Allia Bay, Kanapoi and Mursi, each associated with a different abundance of hominin fossils; respectively c.70%, c.30% and absence. We expected to see a relationship between environments and hominin abundances within the three collections. Instead, we see a resemblance between the localities inhabited by hominins, Allia Bay and Kanapoi and those that are not, namely Mursi.

Results indicate that the range of δ13C values represented in the Mursi dataset is much larger than that of fossils from Allia Bay and Kanapoi. The values extracted from

Mursi fossils are also generally more 13C-depleted, which indicated that the taxa were

incorporating more C3 plants in their diet.

All of the suids in the dataset have low oxygen values, which is expected for obligate drinkers. δ18O values are lowest at Mursi, indicating a close association between suids and the water source. Overall, carbon and oxygen isotopic data from Allia Bay and

Kanapoi suid enamel overlap. We also observe a co-variation of the carbon and oxygen values, as expected.

Surface modification patterns are indicators of the ecological conditions under which the specimens were deposited. There was no detectable pattern in the difference between the collections with respect insect damage, trampling, digestion and root etching.

All three assemblages showed low levels of digestion, which suggests either relatively few carnivores, or they were not stressed to the point of having to extract extra nutrients

136

from bones (Andrews and Fernández-Jalvo 2012; Njau and Blumenschine 2006). As with all bone surface modifications, an alternative explanation is that they were rapidly buried.

There are low frequencies of insect damage and trampling in the three collections, which indicates that none of these localities were frequented by large numbers of herd animals

(Blasco et al. 2008) or osteophagous insects (Fejfar and Kaiser 2005).

Traces left by carnivores are more common at Kanapoi, which is unsurprising considering carnivores skeletal remains are also the most common in that assemblage.

The presence of large Felidae like Homotherium and Dinofelis at Kanapoi is also consistent with more open settings (Lewis 1997; Werdelin and Lewis 2018; Werdelin and

Manthi 2012). Werdelin and Lewis (2018, p.1) aptly describe the Carnivore community at Kanapoi as “ (…) foreshadowing the better known, relatively stable associations of the latest Early Pliocene to Early Pleistocene”.

Levels of surface corrosion were particularly high in the Mursi fossils, suggesting that the fossils submerged in water or buried in acidic sediments, which is consistent with sedimentology at Mursi, which shows evidence of standing water (Drapeau et al. In prep)

(Drapeau et al. In prep). This is consistent with the high levels of polishing at Mursi, which can be interpreted as bones being dragged in water, which suggests increased environmental energy (i.e., hydraulic energy) at the site (Brett and Baird 1986), but not long distance transport (Behrensmeyer et al. 2000). The evenly distributed weathering levels at Kanapoi are consistent with progressive burial, with bones being exposed at the surface for extended periods of time. Bones without any traces of weathering are overrepresented at Mursi, which suggests that burial was likely more rapid than at the two other fossil assemblages studied. This is also consistent with the presence of standing

137

water at the site, since immersion in water precludes surface erosion (Haynes 1988).

Feibel (2011) also notes that tephra influx is a way skeletal remains can be buried rapidly in the context of Plio-Pleistocene eastern Africa sites.

Our results are consistent with the paleoenvironments reconstructed from the sedimentary sequences at each site (Drapeau et al. In prep; Feibel 2011; Wynn 2000). In addition, previous paleoecological studies also suggested that the northern part was more humid than the rest of the Omo-Turkana Basin (Bobe 2011; Drapeau et al. In prep). An ecometric study of dental metrics from Turkana Basin mammals reached similar conclusions and shows for the 4 Ma time period, the most precipitation in the north of the basin (towards the Omo), and the least humid settings in West Turkana (Fortelius et al.

2016). The results of a clumped carbonate isotope study support the claim that hominins from c.4 Ma onwards inhabited hot (soil temperature over 30°C), presumably savannah-like or marginal (i.e. gallery forests) environments in the Turkana Basin

(Passey et al. 2010). Wynn (2004) also describes a decrease in mean annual precipitation during the period between 4.5 and 1.4 Ma. All these results are consistent with hominins inhabiting more open and dry environments when other types of environments are available on the landscape.

However, some of these previous studies reconstruct the paleoenvironments of

Kanapoi as much drier than those of Allia Bay, whereas our results suggest similarities between the two fossil collections. For example, a recent study indicates arid conditions at Kanapoi and possible mesic conditions at Allia Bay (Blumenthal et al. 2017).

Like 4.4 Ma hominid Ardipithecus ramidus (White et al. 2009b), Australopithecus

anamensis exploited C3 resources, whereas later hominins including direct descendant A.

138

afarensis exploited a much larger range of C3-C4 resources (Sponheimer et al. 2013).

The post-canine megadontia, thick enamel, robust mandibles and dental microwear signature of A. anamensis suggest that they likely exploited harder and more abrasive

food within the C3 spectrum (Estebaranz et al. 2012; Macho et al. 2005; Ungar et al.

2018; Ward et al. 2001), which could be found in mesic or dry environments. Inhabiting relatively dry settings, in turn, is consistent with the appearance of obligate bipedal locomotion, plus a tall and slender bauplan, the ability to sweat and the loss of body hair

(Wheeler 1991).

Conclusion

In conclusion, our results demonstrate that the environments at Mursi were different from those of Allia Bay and Kanapoi around 4 million years ago. We found evidence of overlap between the carbon isotopic enamel data between suids from the Kanapoi and

Allia Bay assemblages, whereas the values are 13C-depleted in suids from the Mursi assemblage. In addition, although fossils attributed to mixed-feeding Nyanzachoerus, and specifically Ny. kanamensis, clearly dominate Suidae in the three assemblages, the more grazing Notochoerus is rare at Mursi compared to the other two assemblages. Our analysis of the surface modification patterns on the three collections also supports Mursi being more humid than Allia Bay and Kanapoi: there were higher abrasion and corrosion levels and a larger variety of weathering levels.

In sum, mammal enamel isotopic data, taxonomy and surface modification processes support the presence of more humid settings at Mursi than at Allia Bay and

139

Kanapoi, which could explain the absence of hominins at Mursi. Studies like this one demonstrate the importance of comparing sites where hominins are found with those were they are not. These studies have the potential to reveal aspects of the paleoecology of additional hominin species. Within the Omo-Turkana Basin, A. anamensis seems to have been restricted to the more open and less humid settings found at Allia Bay and

Kanapoi.

Acknowledgements

We thank the Leakey Foundation, Sigma Xi Grants-in-Aid of Research, Explorers Club

Washington Group Inc., Evolving Earth Foundation, Cosmos Club Foundation and the

Lewis N. Cotlow Fund (to LD) for funding this research. We are grateful to the staff of the Ethiopian National Museum in Addis Ababa, Ethiopia and the Nairobi National

Museum in Nairobi, Kenya for access to the fossil collections. We thank the teams who recovered and curated the fossils that comprise the Allia Bay, Kanapoi and Mursi collections.

140

Chapter 5: Conclusion

5.1 Summary

This dissertation examined the paleoenvironments of Australopithecus anamensis in the

Omo-Turkana Basin. I started with a study of the mesowear and hypsodonty of the ungulates from Kanapoi. Results support the presence of heterogeneous, but grassy environments at the site. I then focused on the two most abundant taxa at Allia Bay,

Mursi and Kanapoi: Bovidae and Suidae. I studied bovid remains through a thorough paleontological description of the Allia Bay collection, followed by a community ecology analysis, a description of the mesowear patterns observable in the teeth from the assemblage, a study of the ecomorphology of the astragali and the carbon and oxygen stable isotopic values from mammalian tooth enamel. Taken together, the results from the methods suggest the presence of an ecological gradient from more closed settings at

Mursi, to intermediate settings at Allia Bay to open settings at Kanapoi. The following chapter, which focused on fossil Suidae, provided further information about the paleoecology of each fossil locality, particularly in terms of humidity levels. Aridity is not always linked to vegetation cover (Levin 2015); thus it was important to study both variables separately. Bovids are excellent proxies to infer vegetation because of the within-clade ecological differences. The within-clade diet differences of suids are less pronounced, but they can help provide additional information about humidity levels.

Although Mursi once again stood out as being different, this time the fossil collections from Allia Bay and Kanapoi were largely similar in their taxonomic distribution. The

141

stable isotope values recorded in the tooth enamel and the surface modification patterns highlighted differences between the three collections. These results suggested the presence of more humid settings at Mursi. This conclusion may at least partially explain why Australopithecus anamensis did not seem to have inhabited the locality despite its geographical proximity to known hominin sites.

Our conclusions rest on the abundance of hominins at Kanapoi, Allia Bay and

Mursi as they are currently known. Paleontology is a historical science and, as such, is highly dependent on new discoveries for testing hypotheses, and it is thus possible that new discoveries will change the story our data tell, but there would have to be substantial differences in the types of evidence from the three fossil sites to change our current interpretations.

5.2 Overarching Conclusion

Overall, the analysis of paleoecology through fossil mammals indicate that the site of

Kanapoi, which preserves the most hominin specimens, is also the driest and most open of the three examined. Mursi, where no hominin remains have been found thus far, is more closed and humid. The site of Allia Bay, where fewer hominin remains were found, is in some respects intermediate between the two sites, but much closer to Kanapoi, notably in terms of being drier than Mursi. All in all, it seems that early hominins already favored local environments where the lineage thrived in the rest of the Plio-Pleistocene in

East Africa, as demonstrated by the trend towards an overall expansion of C4 resources and open environments throughout the Plio-Pleistocene documented in the scientific

142

literature (Bobe et al. 2007a; Cerling et al. 2011; Wynn 2004). An alternative explanation is that the Mursi collection was in fact much older than the assemblages from Allia Bay and Kanapoi, and thus may even precede the emergence of A. anamensis. This is, however, unlikely given the correlations between the three faunal assemblages.

These environmental changes are consistent with the suite of adaptations that characterize the hominin lineage including obligate bipedal locomotion, post-canine megadontia, a tall and slender bauplan, the ability to sweat and the loss of body hair

(Wheeler 1991). Our results are consistent with some of the main hypotheses about bipedalism, including the one that stipulates bipedalism emerged as an adaptation to reach for fruits in low branches or in trees (Hunt 1994; Sponheimer et al. 2013). This hypothesis is the most consistent with evidence from carbon stable isotope values of the enamel of A. anamensis suggesting that the hominin had a diet dominated by C3 resources as well as microwear analyses, which suggest similarities between the diet of A. anamensis and that of modern chimpanzees and gorillas (Grine et al. 2006). Studies focusing on the teeth of this hominin can be brought together to suggest that A. anamensis likely exploited hard and abrasive foods within the C3 spectrum. Such evidence includes their post-canine megadontia, thick enamel, robust mandibles and dental microwear signature (Macho et al. 2005; Ungar et al. 2018; Ward et al. 2001).

The findings of this dissertation are important because they impact our knowledge of the earliest known species of the hominin lineage. In addition, sites that preserve c.4

Ma sediments are relatively rare and they contain limited information in comparison to later sites rich in traces of human behaviors including cutmarks and hominin-manufactured tools. For this reason, it is important to analyze these sites as

143

thoroughly as possible. Our results also demonstrate the importance of studying multiple fossil collections together and using multiple lines of evidence in combination, as they all give results that indicate slightly different aspects of the paleoecology of a species.

5.3 Future directions

A comprehensive analysis of all taxa from the three fossil collections was beyond the scope of this thesis, but constitutes our next step. In addition, a comparison of all known

A. anamensis localities (i.e., including data from the Awash Basin) is the next step to understand the paleoecology of that species. In addition, including other methods such as microwear in the inter-locality analysis would provide additional nuances to our results.

144

References Andrews P. 1995. Experiments in taphonomy. J Archaeol Sci 22(2):147-153.

Andrews P. 1996. Palaeoecology and hominoid palaeoenvironments. Biol Rev Camb Philos 71(2):257-300.

Andrews P, and Fernández-Jalvo Y. 2012. How to approach perimortem injury and other modifications. Forensic Microscopy for Skeletal Tissues: Springer. p 191-225.

Andrews P, and Hixson S. 2014. Taxon-Free Methods of Palaeoecology. Ann Zool Fenn 51(1-2):269-284.

Auguste P. 1994. Actions climatiques et épidaphiques: synthèse générale: Actions climatique et édaphique. Artefacts 9:17-27.

Ayliffe LK, and Chivas AR. 1990. Oxygen isotope composition of the bone phosphate of Australian kangaroos: potential as a palaeoenvironmental recorder. Geochim Cosmochim 54(9):2603-2609.

Backwell LR, and d'Errico F. 2001. Evidence of termite foraging by Swartkrans early hominids. Proc Natl A Sci 98(4):1358-1363.

Backwell LR, Parkinson AH, Roberts EM, d'Errico F, and Huchet J-B. 2012. Criteria for identifying bone modification by termites in the fossil record. Palaeogeogr Palaeocl 337:72-87.

Bärmann EV, and Schikora T. 2014. The polyphyly of Neotragus–results from genetic and morphometric analyses. Mamm Biol-Z Säugetierk 79(4):283-286.

Barr WA. 2014. The paleoenvironments of early hominins in the Omo Shungura Formation (Plio-Pleistocene, Ethiopia): synthesizing multiple lines of evidence using phylogenetic ecomorphology: University of Texas at Austin.

Barr WA. 2015. Paleoenvironments of the Shungura Formation (Plio-Pleistocene: Ethiopia) based on ecomorphology of the bovid astragalus. J Hum Evol 88:97- 107.

145

Barton R, Whiten A, Strum S, Byrne R, and Simpson A. 1992. Habitat use and resource availability in baboons. Anim Behav 43(5):831-844.

Beekman J, and Prins H. 1989. Feeding strategies of sedentary large herbivores in East Africa, with emphasis on the African buffalo, Syncerus coffer. Afri J Ecol 27(2):129-147.

Behrensmeyer AK. 1978. Taphonomic and ecologic information from bone weathering. Paleobiol 4(2):150-162.

Behrensmeyer AK. 1982. Time resolution in fluvial vertebrate assemblages. Paleobiol 8(3):211-227.

Behrensmeyer AK. 1984. Taphonomy and the fossil record: the complex processes that preserve organic remains in rocks also leave their own traces, adding another dimension of information to fossil samples. Am Sci 72(6):558-566.

Behrensmeyer AK. 1985. Taphonomy and the paleoecologic reconstruction of hominid habitats in the Koobi Fora Formation. In: Coppens Y, editor. L’Environnement des Hominidés au Plio-Pléistocene. Paris: Masson. p 309-323.

Behrensmeyer AK, and Kidwell SM. 1985. Taphonomy's contributions to paleobiology. Paleobiol 11(1):105-119.

Behrensmeyer AK, Kidwell SM, and Gastaldo RA. 2000. Taphonomy and paleobiology. Paleobiol 26(sp4):103-147.

Behrensmeyer AK, and Reed KE. 2013. Reconstructing the habitats of Australopithecus: Paleoenvironments, site taphonomy, and faunas. The paleobiology of Australopithecus. Dordrecht: Springer Netherlands. p 41-60.

Bibi F. 2011. Tragelaphus nakuae: evolutionary change, biochronology, and turnover in the African Plio-Pleistocene. Zool J Linn Soc Lond 162(3):699-711.

Bibi F, Bukhsianidze M, Gentry AW, Geraads D, Kostopoulos DS, and Vrba ES. 2009. The fossil record and evolution of Bovidae. Palaeontol Electr 12(3):1-11.

146

Bishop L. 1999. Suid paleoecology and habitat preferences at African Pliocene and Pleistocene hominid localities. African Biogeography, Climate Change, and Human Evolution. Oxford: Oxford University Press. p 216-225.

Bishop L. 2010. Suoidea. In: Werdelin L, and Sanders W, editors. Cenozoic Mammals of Africa: University of California Press, Berkeley.

Bishop LC. 1994. Pigs and the ancestors: hominids, suids and environments during the Plio-Pleistocene of East Africa: Yale University.

Bishop LC, King T, Hill A, and Wood B. 2006. Palaeoecology of Kolpochoerus heseloni (= K. limnetes): a multiproxy approach. T Roy Soc S Afr 61(2):81-88.

Bishop LC, Plummer TW, Hertel F, and Kovarovic K. 2011. Paleoenvironments of Laetoli, Tanzania as determined by antelope habitat preferences. Paleontology and geology of Laetoli: Human evolution in context: Springer. p 355-366.

Blasco R, Rosell J, Peris JF, Cáceres I, and Vergès JM. 2008. A new element of trampling: an experimental application on the Level XII faunal record of Bolomor Cave (Valencia, Spain). J Archaeol Sci 35(6):1605-1618.

Blondel C, Merceron G, Andossa L, Taisso MH, Vignaud P, and Brunet M. 2010. Dental mesowear analysis of the late Miocene Bovidae from Toros-Menalla (Chad) and early hominid habitats in Central Africa. Palaeogeogr Palaeocl 292(1):184-191.

Blondel C, Rowan J, Merceron G, Bibi F, Negash E, Barr WA, and Boisserie J-R. 2018. Feeding ecology of Tragelaphini (Bovidae) from the Shungura Formation, Omo Valley, Ethiopia: Contribution of dental wear analyses. Palaeogeogr Palaeocl 496:103-120.

Blumenthal SA, Levin NE, Brown FH, Brugal J-P, Chritz KL, Harris JM, Jehle GE, and Cerling TE. 2017. Aridity and hominin environments. Proc Natl A Sci 114(28):7331-7336.

Bobe R. 1997. Hominid environments in the Pliocene: an analysis of fossil mammals from the Omo Valley, Ethiopia [Dissertation]. Seattle: University of Washington.

Bobe R. 2011. Fossil Mammals and Paleoenvironments in the Omo-Turkana Basin. Evol Anthropol 20(6):254-263.

147

Bobe R, Alemseged Z, and Behrensmeyer AK. 2007a. Approaches to the analysis of faunal change during the East African Pliocene. Hominin environments in the East African Pliocene: an assessment of the faunal evidence: Springer. p 1-24.

Bobe R, and Behrensmeyer AK. 2004. The expansion of grassland ecosystems in Africa in relation to mammalian evolution and the origin of the genus Homo. Palaeogeogr Palaeocl 207(3):399-420.

Bobe R, Behrensmeyer AK, Eck GG, and Harris JM. 2007b. Patterns of abundance and diversity in late Cenozoic bovids from the Turkana and Hadar Basins, Kenya and Ethiopia. Hominin environments in the East African Pliocene: An assessment of the faunal evidence. Dordrecht: Springer. p 129-157.

Bobe R, and Eck GG. 2001. Responses of African bovids to Pliocene climatic change. Paleobiol 27(2):1-48.

Bobe R, and Leakey MG. 2009. Ecology of Plio-Pleistocene mammals in the Omo— Turkana Basin and the emergence of Homo. In: Grine FE, Fleagle JG, and Leakey RE, editors. The First Humans–Origin and Early Evolution of the Genus Homo. Dordrecht: Springer Netherlands. p 173-184.

Boisserie J-R. 2018. Hippopotamidae (Cetartiodactyla, Hippopotamoidea) from Kanapoi, Kenya, and the taxonomic status of the late early Pliocene hippopotamids from the Turkana Basin. J Hum Evol In press.

Boisserie J-R, Souron A, Mackaye HT, Likius A, Vignaud P, and Brunet M. 2014. A new species of Nyanzachoerus (Cetartiodactyla: Suidae) from the late Miocene Toros- Ménalla, Chad, Central Africa. PloS one 9(8):e103221.

Bonnefille R, Potts R, Chalié F, Jolly D, and Peyron O. 2004. High-resolution vegetation and climate change associated with Pliocene Australopithecus afarensis. P Natl Acad Sci 101(33):12125-12129.

Brain CK. 1967. Hottentot food remains and their bearing on the interpretation of fossil bone assemblages. Sci P Namib Desert Res Sta(32):1-11.

Brett CE, and Baird GC. 1986. Comparative taphonomy: a key to paleoenvironmental interpretation based on fossil preservation. Palaios:207-227.

148

Brierley CM, Fedorov AV, Liu Z, Herbert TD, Lawrence KT, and LaRiviere JP. 2009. Greatly expanded tropical warm pool and weakened Hadley circulation in the early Pliocene. Science 323(5922):1714-1718.

Brochu CA. 2018. Pliocene crocodiles from Kanapoi, Turkana Basin, Kenya. J Hum Evol In Press.

Brown F, and Feibel C. 1991. Stratigraphy, depositional environments and palaeogeography of the Koobi Fora Formation. Koobi Fora research project. p 1- 30.

Brunet M, Guy F, Pilbeam D, Mackaye HT, Likius A, Ahounta D, Beauvilain A, Blondel C, Bocherens H, and Boisserie J-R. 2002. A new hominid from the Upper Miocene of Chad, Central Africa. Nature 418(6894):145-151.

Brunet M, and White TD. 2001. Deux nouvelles espèces de Suini (Mammalia, Suidae) du continent africain (Ethiopie; Tchad). Cr Acad Sci II A 332(1):51-57.

Cáceres I, Bennàsar M, Huguet R, Rosell J, Saladié P, Allué E, Solé A, Blasco R, Campeny G, and Esteban-Nadal M. 2012. Taphonomy of level J of Abric Romaní. High Resolution Archaeology and Neanderthal Behavior: Springer. p 159-185.

Cane MA, and Molnar P. 2001. Closing of the Indonesian seaway as a precursor to east African aridification around 3–4 million years ago. Nature 411(6834):157.

Cantalapiedra JL, FitzJohn RG, Kuhn TS, Fernández MH, DeMiguel D, Azanza B, Morales J, and Mooers AØ. 2014. Dietary innovations spurred the diversification of ruminants during the Caenozoic. Proc R Soc B 281(1776):20132746.

Cerling TE, Andanje SA, Blumenthal SA, Brown FH, Chritz KL, Harris JM, Hart JA, Kirera FM, Kaleme P, and Leakey LN. 2015. Dietary changes of large herbivores in the Turkana Basin, Kenya from 4 to 1 Ma. P Natl A Sci 112(37):11467-11472.

Cerling TE, Ehleringer JR, and Harris JM. 1998. Carbon dioxide starvation, the development of C4 ecosystems, and mammalian evolution. Philos T Roy Soc B 353(1365):159-171.

149

Cerling TE, Harris JM, and Passey BH. 2003. Diets of East African Bovidae based on stable isotope analysis. J Mammal 84(2):456-470.

Cerling TE, Manthi FK, Mbua EN, Leakey LN, Leakey MG, Leakey RE, Brown FH, Grine FE, Hart JA, Kaleme P et al. . 2013. Stable isotope-based diet reconstructions of Turkana Basin hominins. P Natl A Sci 110(26):10501-10506.

Cerling TE, Wynn JG, Andanje SA, Bird MI, Korir DK, Levin NE, Mace W, Macharia AN, Quade J, and Remien CH. 2011. Woody cover and hominin environments in the past 6 million years. Nature 476(7358):51-56.

Codron D, Codron J, Lee-thorp A, Sponheimer M, and Brink S. 2006. Dietary variation in impala Aepyceros melampus recorded by carbon isotope composition of feces. Dong wu xue bao[Acta zoologica Sinica] 52(6):1015-1025.

Coffing K, Feibel C, Leakey M, and Walker A. 1994. Four-million-year-old hominids from East Lake Turkana, Kenya. Am J Phys Anthropol 93(1):55-65.

Cohen K, Harper D, and Gibbard P. 2017. ICS International Chronostratigraphic Chart. International Comission on Stratigraphy, IUGS.

Cooke H. 1978. Suid evolution and correlation of African hominid localities: an alternative taxonomy. Science 201(4354):460-463.

Cooke H. 1997. The status of the african fossil suids Kolpochoerus limnetes (Hopwood, 1926), K. phacochoeroides (Thomas 1884) and “K.” afarensis (Cooke, 1978). Geobios 30(1):121-126.

Cooke H, and Maglio V. 1972. Plio-Pleistocene stratigraphy in East Africa in relation to proboscidean and suid evolution. In: Bishop WW, and Miller JA, editors. Calibration of hominoid evolution: Recent Advances in Isotopic and Other Dating Methods as Applicable to Origin of Man. New York: Scottish Academic Press. p 303-329.

Cooke H, and Wilkinson A. 1978. Suidae and tayassuidae. Evol Afr Mammal:435-482.

Coombs MC, and Coombs Jr WP. 1997. Analysis of the geology, fauna, and taphonomy of Morava Ranch Quarry, early Miocene of northwest Nebraska. Palaios:165-187.

150

Coppens Y. 1971. New species of Suidae of Villafranchian of Tunesia, Nyanzachoerus jaegeri nov. sp. CR Acad Sci II D 272(26):3264-&.

Coppens Y, Howell FC, Isaac GL, and Leakey REF, editors. 1976. Earliest Man and environments in the Lake Rudolf Basin: stratigraphy, paleoecology, and evolution: University of Chicago Press.

Curran SC, and Haile-Selassie Y. 2016. Paleoecological reconstruction of hominin- bearing middle Pliocene localities at Woranso-Mille, Ethiopia. J Hum Evol 96:97- 112. d'Errico F, Backwell L, and Berger L. 2001. Bone tool use in termite foraging by early hominids and its impact on our understanding of early hominid behaviour. S Afr J Sci 97(3):71-75.

Damuth J, and Janis CM. 2011. On the relationship between hypsodonty and feeding ecology in ungulate mammals, and its utility in palaeoecology. Biol Rev 86(3):733-758.

Dansgaard W. 1964. Stable isotopes in precipitation. Tellus 16(4):436-468.

Dart RA. 1925. Australopithecus africanus: The Man-Ape of South Africa. Nature 115:195-199.

Davis M, and Pineda Munoz S. 2016. The temporal scale of diet and dietary proxies. Ecol Evol 6(6):1883-1897. de Heinzelin J. 1983. The Omo Group. Musée Royale de l’Afrique Central, Tervuren, Belgique. Annales Série in 8(85).

Delfino M. 2018. Early Pliocene anuran fossils from Kanapoi, Kenya, and the first fossil record for the African burrowing frog Hemisus (Neobatrachia: Hemisotidae). J Hum Evol In press. deMenocal PB. 2004. African climate change and faunal evolution during the Pliocene– Pleistocene. Earth Planet Sci Lett 220(1-2):3-24.

Denys C. 2002. Taphonomy and experimentation. Archaeometry 44(3):469-484.

151

Denzin NK. 2012. Triangulation 2.0. J mix method res 6(2):80-88.

Dowsett H. 2007. The PRISM palaeoclimate reconstruction and Pliocene sea-surface temperature. In: Williams M, Haywood A, Gregory F, and Schmidt D, editors. Deep-time perspectives on climate change: marrying the signal from computer models and biological proxies. London: The Geological Society for the Micropalaeontological Society. p 459-480.

Dowsett HJ, Barron JA, Poore RZ, Thompson RS, Cronin TM, Ishman SE, and Willard DA. 1999. Middle Pliocene paleoenvironmental reconstruction: PRISM2. US Geological Survey open file report 99:535.

Drapeau MS, Bobe R, Geraads D, Wynn JG, and Dumouchel L. In prep. Paleontology and taphonomy of the Mursi Formation. In: Bobe R, and Reynolds S, editors. Pliocene and Early Pleistocene African Fauna and Paleoecology: Cambridge University Press.

Drapeau MS, Bobe R, Wynn JG, Campisano CJ, Dumouchel L, and Geraads D. 2014. The Omo Mursi Formation: A window into the East African Pliocene. J Hum Evol 75:64-79.

Dumouchel L, and Bobe R. In review. Paleoecological implications of dental mesowear and hypsodonty in fossil ungulates from Kanapoi J Hum Evol.

Eberth DA, Rogers RR, and Fiorillo AR. 2007. A practical approach to the study of bonebeds. Bonebeds: genesis, analysis, and paleobiological significance. p 265- 331.

Efremov IA. 1940. Taphonomy: a new branch of paleontology. Pan-Am Geol 74(2):81- 93.

Eronen J, Puolamäki K, Liu L, Lintulaakso K, Damuth J, Janis C, and Fortelius M. 2010. Precipitation and large herbivorous mammals II: application to fossil data. Evol Ecol Res 12(2):235-248.

Estebaranz F, Galbany J, Martínez LM, Turbón D, and Pérez-Pérez A. 2012. Buccal dental microwear analyses support greater specialization in consumption of hard foodstuffs for Australopithecus anamensis. J Archaeol Sci 90:163-185.

152

Everett MA. 2010. The paleoecology of the Pleistocene Upper Busidima Formation, Gona, Afar Depression, Ethiopia: Indiana University.

Faith JT, Choiniere JN, Tryon CA, Peppe DJ, and Fox DL. 2011. Taxonomic status and paleoecology of Rusingoryx atopocranion (Mammalia, Artiodactyla), an extinct Pleistocene bovid from Rusinga Island, Kenya. Quaternary Research 75(3):697- 707.

Feibel C. 2003. Stratigraphy and depositional setting of the Pliocene Kanapoi Formation, lower Kerio valley, Kenya. Contrib Sci 498:9-20.

Feibel CS. 1988. Paleoenvironments of the Koobi Fora Formation, Turkana Basin, Northern Kenya: Department of Geology and Geophysics, University of Utah.

Feibel CS. 2011. A geological history of the Turkana Basin. Evol Anthropol 20(6):206- 216.

Fejfar O, and Kaiser TM. 2005. Insect bone-modification and paleoecology of Oligocene mammal-bearing sites in the Doupov Mountains, northwestern Bohemia. Palaeontol Electron 8(1):1.

Feranec RS. 2003. Stable isotopes, hypsodonty, and the paleodiet of Hemiauchenia (Mammalia: Camelidae): a morphological specialization creating ecological generalization. Paleobiol 29(2):230-242.

Fernandez-Jalvo Y, and Andrews P. 2016. Atlas of taphonomic identifications: 1001+ images of fossil and recent mammal bone modification: Springer.

Fesseha N. 1999. Systematics of Hadar (Afar, Ethiopia) Suidae. Washington D.C.: Howard University.

Field DJ. 2018. Preliminary paleoecological insights from the Pliocene avifauna of Kanapoi, Kenya: Implications for the ecology of Australopithecus anamensis. J Hum Evol In press.

Fiorillo AR. 1988. Taphonomy of Hazard Homestead Quarry (Ogallala Group), Hitchcock County, Nebraska. Rocky Mt Geol 26(2):57-97.

153

Fiorillo AR. 1989. An experimental study of trampling: implications for the fossil record. Bone modification. Orono, Maine: Center for the study of the first Americans. p 61-71.

Forrest FL, Plummer TW, and Raaum RL. 2018. Ecomorphological analysis of bovid mandibles from Laetoli Tanzania using 3D geometric morphometrics: Implications for hominin paleoenvironmental reconstruction. J Hum Evol 114:20- 34.

Fortelius M, Eronen J, Liu L, Pushkina D, Tesakov A, Vislobokova I, and Zhang Z. 2006. Late Miocene and Pliocene large land mammals and climatic changes in Eurasia. Palaeogeogr Palaeocl 238(1-4):219-227.

Fortelius M, and Solounias N. 2000. Functional characterization of ungulate molars using the abrasion-attrition wear gradient: a new method for reconstructing paleodiets. Am Mus Nov 3301:1-36.

Fortelius M, Žliobaitė I, Kaya F, Bibi F, Bobe R, Leakey L, Leakey M, Patterson D, Rannikko J, and Werdelin L. 2016. An ecometric analysis of the fossil mammal record of the Turkana Basin. Phil Trans R Soc B 371(1698):20150232.

Franz-Odendaal TA, and Kaiser TM. 2003. Differential mesowear in the maxillary and mandibular cheek dentition of some ruminants (Artiodactyla). Ann Zool Fennici: JSTOR. p 395-410.

Fraser D, and Theodor JM. 2011. Comparing ungulate dietary proxies using discriminant function analysis. J Morphol 272(12):1513-1526.

Gagnon M, and Chew AE. 2000. Dietary preferences in extant African Bovidae. J Mammal 81(2):490-511.

Gentry AW, and Gentry A. 1978. Fossil Bovidae (Mammalia) of Olduvai Gorge, Tanzania. Bull Brit Mus (Nat Hist): Geol 29.

Geraads D. 2004. New skulls of Kolpochoerus phacochoeroides (Suidae, Mammalia) from the late Pliocene of Ahl al Oughlam, Morocco. Palaeontol Afr 40:69-83.

Geraads D. 2018. Perissodactyla (Rhinocerotidae and Equidae) from Kanapoi. J Hum Evol In press.

154

Geraads D, and Bobe R. 2018a. Ruminants (Giraffidae and Bovidae) from Kanapoi. J Hum Evol In press.

Geraads D, and Bobe R. 2018b. Suidae from Kanapoi. J Hum Evol In press.

Geraads D, Bobe R, and Manthi FK. 2013. New ruminants (Mammalia) from the Pliocene of Kanapoi, Kenya, and a revision of previous collections, with a note on the Suidae. J Afr Earth Sci 85:53-61.

Geraads D, Melillo S, and Haile-Selassie Y. 2009. Middle Pliocene Bovidae from hominid-bearing sites in the Woranso-Mille area, Afar region, Ethiopia. Palaeontol Afr 44:59-70.

Geraads D, Raynal J-P, and Eisenmann V. 2004. The earliest human occupation of North Africa: a reply to Sahnouni et al.(2002). J Hum Evol 46:751-761.

Gifford D. 1991. Bones are not enough: analogies, knowledge, and interpretative strategies in zooarcheology. J Anthropol Archaeol 10:215-254.

Green DJ, Gordon AD, and Richmond BG. 2007. Limb-size proportions in Australopithecus afarensis and Australopithecus africanus. J Hum Evol 52(2):187-200.

Greenacre MJ, and Vrba ES. 1984. Graphical display and interpretation of antelope census data in African wildlife areas, using correspondence analysis. Ecol Soc Am 65(3):984-997.

Grine FE, Ungar PS, and Teaford MF. 2006. Was the Early Pliocene hominin 'Australopithecus' anamensis a hard object feeder? S Afr J Sci 102(7-8):301-310.

Grubb P. 2001. Review of family-group names of living bovids. J Mammal 82(2):374- 388.

Guillemot C. 1997. Recherches paléoanthropologiques dans la basse vallée de l’Omo. Bull maison étud éthiop 11:69-78.

Hagemann SI. 2010. Paleoecology and taphonomy of a hominid-bearing site: locality 261-1, Allia Bay, Kenya: Rutgers University-Graduate School-New Brunswick.

155

Haile-Selassie Y, and Simpson SW. 2013. A new species of Kolpochoerus (Mammalia: Suidae) from the Pliocene of Central Afar, Ethiopia: its taxonomy and phylogenetic relationships. J Mammal Evol 20(2):115-127.

Haile-Selassie Y, Vrba E, and Bibi F. 2009. Suidae. Ardipithecus kadabba: Late Miocene Evidence from the Middle Awash, Ethiopia. Berkeley: University of California Press. p 331-371.

Haileab B, Brown FH, McDougall I, and Gathogo PN. 2004. Gombe Group basalts and initiation of Pliocene deposition in the Turkana depression, northern Kenya and southern Ethiopia. Geol Mag 141(1):41-53.

Harcourt-Smith WE. 2007. The origins of bipedal locomotion. In: Tattersall I, and Henke W, editors. Handbook of paleoanthropology. Berlin: Springer-Verlag. p 1483- 1518.

Harris J, and Cerling T. 2002. Dietary adaptations of extant and Neogene African suids. J Zool 256(01):45-54.

Harris J, Leakey M, Cerling T, and Winkler A. 2003. Early Pliocene tetrapod remains from Kanapoi, Lake Turkana Basin, Kenya. Contrib Sci 498:39-1114.

Harris J, and Liu L. 2007. Superfamily Suoidea. In: Prothero DR, and Foss SE, editors. The evolution of Artiodactyla. Baltimore: Johns Hopkins University Press. p 130- 150.

Harris JM. 1983a. Background to the Study of the Koobi Fora Fossil Fauna. In: Harris JM, editor. Koobi Fora research project: Volume II: The fossil ungulates, Proboscidea, Perissodactyla, and Suidae Oxford: Oxford University Press.

Harris JM. 1983b. Family Suidae. In: Harris JM, editor. Koobi Fora research project: Volume II: The fossil ungulates, Proboscidea, Perissodactyla, and Suidae. Oxford: Oxford University Press.

Harris JM. 1991. Koobi Fora Research Project: The Fossil Ungulates: Geology, Fossil Artiodactyls, and Palaeoenvironments. New York.

156

Harris JM, Leakey MG, and Brown FH. 1988. Stratigraphy and paleontology of Pliocene and Pleistocene localities west of Lake Turkana, Kenya: Natural History Museum of Los Angeles County.

Harris JM, and White TD. 1979. Evolution of the Plio-Pleistocene African Suidae. Trans Am Phil Soc 69(2):1-128.

Harrison T. 2010. Apes among the tangled branches of human origins. Science 327(5965):532-534.

Hassanin A, Delsuc F, Ropiquet A, Hammer C, van Vuuren BJ, Matthee C, Ruiz-Garcia M, Catzeflis F, Areskoug V, and Nguyen TT. 2012. Pattern and timing of diversification of Cetartiodactyla (Mammalia, Laurasiatheria), as revealed by a comprehensive analysis of mitochondrial genomes. CR Biol 335(1):32-50.

Hatley T, and Kappelman J. 1980. Bears, pigs, and Plio-Pleistocene hominids: a case for the exploitation of belowground food resources. Hum Ecol 8(4):371-387.

Havemann CP, Retief TA, Tosh CA, and Bruyn P. 2016. Roan antelope Hippotragus equinus in Africa: a review of abundance, threats and ecology. Mamm Rev 46(2):144-158.

Haynes G. 1988. Longitudinal studies of African elephant death and bone deposits. J Archeol Sc 15(2):131-157.

Hendey QB, and Cooke HBS. 1985. Kolpochoerus paiceae (Mammalia, Suidae) from Skurwerug near Saldanha, South Africa, and its palaeoenvironmental implications. Ann S Afr Mus 97(2):9-56.

Hunermann KA. 1968. Die Suidae (Mammalia, Artiodactyla) aus den Dinotheriensanden (Unterpliozan= Pont) Rheinhessens (Sud-westdeutschland). Schweiz Palaontol Abh 86:1-96.

Hunt KD. 1994. The evolution of human bipedality: ecology and functional morphology. J Hum Evol 26(3):183-202.

Hunt RM. 1978. Depositional setting of a Miocene mammal assemblage, Sioux County, Nebraska (USA). Palaeogeogr palaeocl 24(1):1-52.

157

IUCN. 2017. The IUCN Red List of Threatened species. Version 2017-3.

Jablonski N, Leakey M, Ward C, and Antón M. 2008. Koobi Fora research project, Vol. 6: The fossil monkeys. San Francisco: California Academy of Sciences.

Janis CM. 1988. An estimation of tooth volume and hypsodonty indices in ungulate mammals, and the correlation of these factors with dietary preference. Teeth revisited: proceedings of the VIIth international symposium on dental morphology: Mémoires du Muséum national d'histoire Naturelle Paris (serie C). p 367-387.

Jernvall J, and Fortelius M. 2002. Common mammals drive the evolutionary increase of hypsodonty in the Neogene. Nature 417(6888):538-540.

Jones DB, and Desantis LR. 2017. Dietary ecology of ungulates from the La Brea tar pits in southern California: A multi-proxy approach. Palaeogeogr palaeocl 466:110- 127.

Kaiser TM. 2011. Feeding ecology and niche partitioning of the Laetoli ungulate faunas. In: Harrison T, editor. Paleontology and Geology of Laetoli: Human evolution in context. Dordrecht: Springer Netherlands. p 329-354.

Kaiser TM, Brasch J, Castell JC, Schulz E, and Clauss M. 2009. Tooth wear in captive wild ruminant species differs from that of free-ranging conspecifics. Mamm Biol- Z Säugetierk 74(6):425-437.

Kaiser TM, Müller DW, Fortelius M, Schulz E, Codron D, and Clauss M. 2013. Hypsodonty and tooth facet development in relation to diet and habitat in herbivorous ungulates: implications for understanding tooth wear. Mammal Rev 43(1):34-46.

Kaiser TM, Solounias N, Fortelius M, Bernor RL, and Schrenk F. 2000. Tooth mesowear analysis on hippotherium primigenium from the Vallesian Dinotheriensande (Germany). Carolinea: Beiträge zur naturkundlichen Forschung in Südwestdeutschland 58:103-114.

Kappelman J. 1988. Morphology and locomotor adaptations of the bovid femur in relation to habitat. J Morph 198(1):119-130.

158

Kappelman J, Swisher III CC, Fleagle JG, Yirga S, Bown TM, and Feseha M. 1996. Age of Australopithecus afarensis from Fejej, Ethiopia. J Hum Evol 30(2):139-146.

Kingdon J, Happold D, Butynski T, Hoffmann M, Happold M, and Kalina JE. 2013a. Carnivores, Pangolins, Equids and Rhinoceroses: Bloomsbury.

Kingdon J, Happold D, Butynski T, Hoffmann M, Happold M, and Kalina JE. 2013b. Pigs, Hippopotamuses, Chevrotain, Giraffes, Deer and Bovids. London: Bloomsbury Publishing.

Kingston JD. 2007. Shifting adaptive landscapes: progress and challenges in reconstructing early hominid environments. YearB Phys Anthropol 50:20-58.

Kingston JD, and Harrison T. 2007. Isotopic dietary reconstructions of Pliocene herbivores at Laetoli: Implications for early hominin paleoecology. Palaeogeogr palaeocl 243(3):272-306.

Kitching JW. 1963. Bone, tooth & horn tools of palaeolithic man: an account of the osteodontokeratic discoveries in Pin Hole Cave, Derbyshire. Manchester, England: Manchester University Press.

Koch PL. 1998. Isotopic reconstruction of past continental environments. Ann Rev Earth Pl Sc 26(1):573-613.

Kolska Horwitz L. 1990. The origin of partially digested bones recovered from archaeological contexts in Israel. Paléorient:97-106.

Kovarovic K, and Andrews P. 2007. Bovid postcranial ecomorphological survey of the Laetoli paleoenvironment. J Hum Evol 52(6):663-680.

Kozma EE, Webb NM, Harcourt-Smith WEH, Raichlen DA, D'Août K, Brown MH, Finestone EM, Ross SR, Aerts P, and Pontzer H. 2018. Hip extensor mechanics and the evolution of walking and climbing capabilities in humans, apes, and fossil hominins. Proc Natl A Sci.

Krishnamurthy R, and Bhattacharya S. 1991. Stable oxygen and hydrogen isotope ratios in shallow ground waters from India and a study of the role of evapotranspiration in the Indian monsoon. Rev Mineral Geochem 3:187-193.

159

Kullmer O. 1999. Evolution of African Plio-Pleistocene suids (Artiodactyla: Suidae) based on tooth pattern analysis. Kaupia 9:1-34.

Kullmer O, Sandrock O, Viola TB, Hujer W, Said H, and Seidler H. 2008. Suids, elephantoids, paleochronology, and paleoecology of the Pliocene hominid site Galili, Somali Region, Ethiopia. Palaios 23(7):452-464.

Lamarck J. 1809. Philosophie zoologique Paris: Duminil-Lesueur.

Leakey LSB. 1958. Some East African Pleistocene Suidae. Fossil Mammals of Africa. London: British Museum. p 1-133.

Leakey MG, Feibel CS, Bernor RL, Harris JM, Cerling TE, Stewart KM, Storrs GW, Walker A, Werdelin L, and Winkler AJ. 1996. Lothagam: a record of faunal change in the Late Miocene of East Africa. J Vert Paleontol 16(3):556-570.

Leakey MG, Feibel CS, McDougall I, and Walker A. 1995. New four-million-year-old hominid species from Kanapoi and Allia Bay, Kenya. Nature 376(6541):565-571.

Leakey MG, Feibel CS, McDougall I, Ward C, and Walker A. 1998. New specimens and confirmation of an early age for Australopithecus anamensis. Nature 393(6680):62-66.

Leakey MG, and Leakey RE. 1978. Koobi Fora research project. Oxford, England and New York, New Yok: Clarendon Press.

Lee-Thorp J, and Sponheimer M. 2015. Contribution of stable light isotopes to paleoenvironmental reconstruction. Handb paleoanthropol: Springer. p 441-464.

Lee-Thorp JA, Sealy JC, and Van Der Merwe NJ. 1989. Stable carbon isotope ratio differences between bone collagen and bone apatite, and their relationship to diet. J Archaeol Sci 16(6):585-599.

Leeder MR. 2009. Sedimentology and sedimentary basins: from turbulence to tectonics. Hoboken: John Wiley & Sons. 608 p.

Levin NE. 2008. Isotopic records of Plio-Pleistocene climate and environments in eastern Africa [Dissertation]: The University of Utah.

160

Levin NE. 2015. Environment and climate of early human evolution. Ann Rev Earth Plan Sc 43.

Levin NE, Brown FH, Behrensmeyer AK, Bobe R, and Cerling TE. 2011. Paleosol carbonates from the Omo Group: Isotopic records of local and regional environmental change in East Africa. Palaeogeogr Palaeocl 307(1):75-89.

Levin NE, Cerling TE, Passey BH, Harris JM, and Ehleringer JR. 2006. A stable isotope aridity index for terrestrial environments. Proc Natl A Sci 103(30):11201-11205.

Lewis ME. 1997. Carnivoran paleoguilds of Africa: implications for hominid food procurement strategies. J Hum Evol 32(2-3):257-288.

Louys J, Ditchfield P, Meloro C, Elton S, and Bishop LC. 2012. Stable isotopes provide independent support for the use of mesowear variables for inferring diets in African antelopes. Proc Roy Soc B: Biol Sci 279(1746):4441-4446.

Louys J, Meloro C, Elton S, Ditchfield P, and Bishop LC. 2011. Mesowear as a means of determining diets in African antelopes. J Archaeol Sci 38(7):1485-1495.

Louys J, Meloro C, Elton S, Ditchfield P, and Bishop LC. 2015. The potential and pitfalls of using simple dental metrics to infer the diets of African antelopes (Mammalia: Bovidae). Palaeontol Afr 49:8-24.

Lovejoy CO. 1981. The origin of man. Science 211(4480):341-350.

Lüttge U. 2004. Ecophysiology of crassulacean acid metabolism (CAM). Ann Bot 93(6):629-652.

Luz B, Kolodny Y, and Horowitz M. 1984. Fractionation of oxygen isotopes between mammalian bone-phosphate and environmental drinking water. Geochim Cosmochim Ac 48(8):1689-1693.

Lyman RL. 1994. Vertebrate taphonomy. Cambridge: Cambridge University Press.

Lyman RL, Houghton LE, and Chambers AL. 1992. The effect of structural density on marmot skeletal part representation in archaeological sites. J Archeol Sc 19(5):557-573.

161

Macho GA. 2014. An ecological and behavioural approach to hominin evolution during the Pliocene. Quaternary Sci Rev 96:23-31.

Macho GA, Shimizu D, Jiang Y, and Spears IR. 2005. Australopithecus anamensis: A finite-element approach to studying the functional adaptations of extinct hominins. Anat Rec A 283(2):310-318.

Magurran AE. 1988. Diversity indices and species abundance models. Ecological diversity and its measurement. Dordrecht: Springer Netherlands. p 7-45.

Manthi FK. 2006. The Pliocene micromammalian fauna from Kanapoi, northwestern Kenya, and its contribution to understanding the environment of Australopithecus anamensis: University of Cape Town.

Manthi FK. 2008. The taphonomy of the Pliocene microfauna from Kanapoi, north- western Kenya. J Taphonomy 6(1):41-67.

Manthi FK, Cerling TE, Chritz KL, and Blumenthal SA. 2018. Diets of mammalian fossil fauna from Kanapoi, northwestern Kenya. J Hum Evol.

Marean CW, Anderson RJ, Bar-Matthews M, Braun K, Cawthra HC, Cowling RM, Engelbrecht F, Esler KJ, Fisher E, and Franklin J. 2015. A new research strategy for integrating studies of paleoclimate, paleoenvironment, and paleoanthropology. Evol Anthropol 24(2):62-72.

Marean CW, and Spencer LM. 1991. Impact of carnivore ravaging on zooarchaeological measures of element abundance. Am Antiquity 56(4):645-658.

Marshall AJ, and Wrangham RW. 2007. Evolutionary consequences of fallback foods. Int J Primatol 28(6):1219.

Matthee CA, and Robinson TJ. 1999. CytochromebPhylogeny of the family Bovidae: resolution within the alcelaphini, antilopini, neotragini, and tragelaphini. Mol Phylogenet Evol 12(1):31-46.

Mendoza M, and Palmqvist P. 2008. Hypsodonty in ungulates: an adaptation for grass consumption or for foraging in open habitat? J Zool 274(2):134-142.

162

Merceron G, Costeur L, Maridet O, Ramdarshan A, and Göhlich UB. 2012. Multi-proxy approach detects heterogeneous habitats for primates during the Miocene climatic optimum in Central Europe. J Hum Evol 63(1):150-161.

Mohr P. 1991. The discovery of African Rift geology: A summary. Magmatism in Extensional Structural Settings: Springer. p 11-22.

Montalvo CI, Pessino ME, and González VH. 2007. Taphonomic analysis of remains of mammals eaten by pumas (Puma concolor Carnivora, Felidae) in central Argentina. J Archeol Sc 34(12):2151-2160.

Ngugi R, Kilonzo J, Kimeu J, and Mureithi S. 2014. Seasonal botanical characteristics of the diets of Grants (Gazella granti Brooke) and Thomsons (Gazella Thomsoni Guenther) in the dryland habitats of south-central, Kenya. Int J of Biodivers Conserv 6(8):581-588.

Njau JK, and Blumenschine RJ. 2006. A diagnosis of crocodile feeding traces on larger mammal bone, with fossil examples from the Plio-Pleistocene Olduvai Basin, Tanzania. J Hum Evol 50(2):142-162.

Olsen SL, and Shipman P. 1988. Surface modification on bone: trampling versus butchery. J Archeol Sc 15(5):535-553.

Pante MC, Muttart MV, Keevil TL, Blumenschine RJ, Njau JK, and Merritt SR. 2017. A new high-resolution 3-D quantitative method for identifying bone surface modifications with implications for the Early Stone Age archaeological record. J Hum Evol 102:1-11.

Passey BH, Levin NE, Cerling TE, Brown FH, and Eiler JM. 2010. High-temperature environments of human evolution in East Africa based on bond ordering in paleosol carbonates. Proc Natl A Sci 107(25):11245-11249.

Patterson B. 1966. A new locality for early Pleistocene fossils in north-western Kenya. Nature 212(5062):577.

Patterson B, Behrensmeyer AK, and Sill WD. 1970. Geology and fauna of a new Pliocene locality in north-western Kenya. Nature 226(5249):918.

163

Patterson B, and Howells WW. 1967. Hominid humeral fragment from early Pleistocene of northwestern Kenya. Science 156(3771):64-66.

Pawłowski B, and Nowaczewska W. 2015. Origins of Homininae and putative selection pressures acting on the early hominins In: Henke W, and Tattersall I, editors. HandB Paleoanthropol. Verlag: Springer Berlin Heidelberg. p 1887-1918.

Pérez-Barbería FJ, Gordon IJ, and Nores C. 2001. Evolutionary transitions among feeding styles and habitats in ungulates. Evol Ecol Res 3(2):221-230.

Peters CR, and Blumenschine RJ. 1995. Landscape perspectives on possible land use patterns for Early Pleistocene hominids in the Olduvai Basin, Tanzania. J Hum Evol 29(4):321-362.

Phoca-Cosmetatou N. 2005. Bone weathering and food procurement strategies: assessing the reliability of our behavioural inferences. In: O'Connor T, editor. Biosphere to lithosphere: new studies in vertebrate taphonomy. Oxford: Oxbow Books. p 135- 145.

Pickford M. 1986. A revision of the Miocene Suidae and Tayassuidae,(Artiodactyla, Mammalia) of Africa. Leiden: Brill Academic Pub.

Pickford M. 1989. New specimens of Nyanzachoerus waylandi (Mammalia, Suidae, Tetraconodontinae) from the type area, Nyaburogo,(Ppper Miocene), Lake Albert Rift, Uganda. Geobios 22(5):641-651.

Pickford M. 2014. Sus Valentini Filhol (1882) from St Gaudens (MN 8-9) France: blighted from the outset but a key to understanding late Middle Miocene Tetraconodontinae (Suidae, Mammalia) of Europe. Mainzer Nat Archiv 51:167- 220.

Plummer TW, and Bishop LC. 1994. Hominid paleoecology at Olduvai Gorge, Tanzania as indicated by antelope remains. J Hum Evol 27(1-3):47-75.

Plummer TW, Bishop LC, and Hertel F. 2008. Habitat preference of extant African bovids based on astragalus morphology: operationalizing ecomorphology for palaeoenvironmental reconstruction. J Archeol Sc 35(11):3016-3027.

164

Plummer TW, Ferraro JV, Louys J, Hertel F, Alemseged Z, Bobe R, and Bishop L. 2015. Bovid ecomorphology and hominin paleoenvironments of the Shungura Formation, lower Omo River Valley, Ethiopia. J Hum Evol 88:108-126.

Pobiner B. 2008. Paleoecological information in predator tooth marks. J Taphonomy 6(3):373-397.

Potts R. 1988. Early hominid activities at Olduvai: Foundations of human behaviour. New York: Aldine De Gruyer.

Rannikko J, Žliobaitė I, and Fortelius M. 2017. Relative abundances and palaeoecology of four suid genera in the Turkana Basin, Kenya, during the late Miocene to Pleistocene. Palaeogeogr palaeocl 487:187-193.

Reda H. 2011. The origin and evolution of Notochoerus euilus (Suidae): fossil evidence from Woranso-Mille, central Afar, Ethiopia: MSc Thesis, Addis Ababa University, Addis Ababa Google Scholar.

Reda HG, Lazagabaster IA, and Haile-Selassie Y. 2017. Newly Discovered Crania of Nyanzachoerus jaegeri (Tetraconodontinae, Suidae, Mammalia) from the Woranso-Mille (Ethiopia) and Reappraisal of Its Generic Status. J Mammal Evol:1-21.

Reed KE. 1997. Early hominid evolution and ecological change through the African Plio- Pleistocene. J Hum Evol 32(2):289-322.

Reed KE. 2013. Multiproxy paleoecology. In: Begun DR, editor. A companion to paleoanthropology: Blackwell Publishing Ltd. p 203-225.

Reitz EJ, and Wing ES. 1999. Zooarchaeology. Cambridge: Cambridge University Press.

Reynolds SC, Bailey GN, and King GC. 2011. Landscapes and their relation to hominin habitats: case studies from Australopithecus sites in eastern and southern Africa. J Hum Evol 60(3):281-298.

Ring U, Albrecht C, and Schrenk F. 2018. The East African Rift System: Tectonics, climate and biodiversity. In: Hoorn C, Perrigo A, and Antonelli A, editors. Mountains, Climate and Biodiversity. Oxford: Wiley Blackwell. p 391.

165

Rivals F, Solounias N, and Mihlbachler MC. 2007. Evidence for geographic variation in the diets of late Pleistocene and early Holocene Bison in North America, and differences from the diets of recent Bison. Quatern Res 68(3):338-346.

Rovero F, and Struhsaker TT. 2007. Vegetative predictors of primate abundance: utility and limitations of a fine-scale analysis. Am J Primatol 69(11):1242-1256.

Rowan J, Faith JT, Gebru Y, and Fleagle JG. 2015. Taxonomy and paleoecology of fossil Bovidae (Mammalia, Artiodactyla) from the Kibish Formation, southern Ethiopia: Implications for dietary change, biogeography, and the structure of the living bovid faunas of East Africa. Palaeogeogr palaeocl 420:210-222.

Sánchez-Hernández C, Rivals F, Blasco R, and Rosell J. 2016. Tale of two timescales: combining tooth wear methods with different temporal resolutions to detect seasonality of Palaeolithic hominin occupational patterns. J Archeol Sc 6:790- 797.

Schoeninger MJ, Reeser H, and Hallin K. 2003. Paleoenvironment of Australopithecus anamensis at Allia Bay, East Turkana, Kenya: evidence from mammalian herbivore enamel stable isotopes. J Archaeol Anthropol 22(3):200-207.

Schubert BW. 2007. Dental mesowear and the palaeodiets of bovids from Makapansgat Limeworks Cave, South Africa. Palaeogeogr palaeocl 241:301-319.

Seki O, Foster GL, Schmidt DN, Mackensen A, Kawamura K, and Pancost RD. 2010. Alkenone and boron-based Pliocene pCO 2 records. Earth Planet Sci Lett 292(1):201-211.

Semaw S, Simpson SW, Quade J, Renne PR, Butler RF, McIntosh WC, Levin N, Dominguez-Rodrigo M, and Rogers MJ. 2005. Early Pliocene hominids from Gona, Ethiopia. Nature 433(7023):301-305.

Senut B, Pickford M, Gommery D, Mein P, Cheboi K, and Coppens Y. 2001. First hominid from the Miocene (Lukeino formation, Kenya). Cr Acad Sci II A 332(2):137-144.

Sepulchre P, Ramstein G, Fluteau F, Schuster M, Tiercelin J-J, and Brunet M. 2006. Tectonic uplift and eastern Africa aridification. Science 313(5792):1419-1423.

166

Sewell L, Merceron G, Hopley P, Zipfel B, and Reynolds SC. 2018. Using springbok (Antidorcas) dietary proxies to reconstruct inferred palaeovegetational changes over 2 million years in Southern Africa. J Archeol Sc.

Shipman P. 1981. Life history of a fossil: an introduction to taphonomy and paleoecology. Cambridge, Massachusetts and London, England: Harvard Univeristy Press.

Shipman P, and Harris JM. 1988. Habitat preference and paleoecology of Australopithecus boisei in eastern Africa. In: Grine FE, editor. Evolutionary history of the “robust” australopithecines: Transaction Publisher. p 343-381.

Solounias N, and Moelleken SM. 1993. Dietary adaptation of some extinct ruminants determined by premaxillary shape. J Mammal 74(4):1059-1971.

Souron A, Boisserie J-R, and White TD. 2013. A new species of the suid genus Kolpochoerus from Ethiopia. Acta Palaeontol Pol 60(1):79-96.

Spencer LM. 1997. Dietary adaptations of Plio-Pleistocene Bovidae: implications for hominid habitat use. J Hum Evol 32(2):201-228.

Sponheimer M, Alemseged Z, Cerling TE, Grine FE, Kimbel WH, Leakey MG, Lee- Thorp JA, Manthi FK, Reed KE, and Wood BA. 2013. Isotopic evidence of early hominin diets. P Natl A Sci 110(26):10513-10518.

Sponheimer M, and Lee-Thorp JA. 1999. Oxygen isotopes in enamel carbonate and their ecological significance. J Archeol Sc 26(6):723-728.

Stewart K. 2003. Fossil fish remains from the Pliocene Kanapoi site, Kenya. Contrib Sci 498:21-38.

Su DF, and Harrison T. 2008. Ecological implications of the relative rarity of fossil hominins at Laetoli. J Hum Evol 55(4):672-681.

Sutcliffe AJ. 1970. Spotted hyaena: crusher, gnawer, digester and collector of bones. Nature 227(5263):1110.

167

Tappen N. 1969. The relationship of weathering cracks to split-line orientation in bone. Am J Phys Anthropol 31(2):191-197.

Thompson C, Ball S, Thompson T, and Gowland R. 2011. The abrasion of modern and archaeological bones by mobile sediments: the importance of transport modes. J Archeol Sc 38(4):784-793.

Thurmond VA. 2001. The point of triangulation. J nurs scholar 33(3):253-258.

Tjelldén AKE, Kristiansen SM, Matthiesen H, and Pedersen O. 2015. Impact of roots and rhizomes on wetland archaeology: a review. Conserv Manage Archa 17(4):370- 391.

Trauth MH, Maslin MA, Deino AL, Junginger A, Lesoloyia M, Odada EO, Olago DO, Olaka LA, Strecker MR, and Tiedemann R. 2010. Human evolution in a variable environment: the amplifier lakes of Eastern Africa. Quaternary Sci Rev 29(23- 24):2981-2988.

Ungar PS. 2015. Mammalian dental function and wear: A review. Biosurf Biotribol 1(1):25-41.

Ungar PS, Abella EF, Burgman JH, Lazagabaster IA, Scott JR, Delezene LK, Manthi FK, Plavcan JM, and Ward CV. 2018. Dental microwear and Pliocene paleocommunity ecology of bovids, primates, rodents, and suids at Kanapoi. J Hum Evol In press.

Uno KT, Cerling TE, Harris JM, Kunimatsu Y, Leakey MG, Nakatsukasa M, and Nakaya H. 2011. Late Miocene to Pliocene carbon isotope record of differential diet change among East African herbivores. P Natl A Sci 108(16):6509-6514.

Van Bocxlaer B. 2018. Paleoecological insights from fossil freshwater mollusks of the Kanapoi Formation (Omo-Turkana Basin, Kenya). J Hum Evol.

Verhaegen M, and Puech P-F. 2000. Hominid lifestyle and diet reconsidered: paleo- environmental and comparative data. J Hum Evol 15(3-4):175-186.

Vignaud P, Duringer P, Mackaye HT, Likius A, Blondel C, Boisserie J-R, de Bonis L, Eisenmann V, Etienne M-E, and Geraads D. 2002. Geology and palaeontology of

168

the Upper Miocene Toros-Menalla hominid locality, Chad. Nature 418(6894):152-155.

Vorrhies M. 1969. Taphonomy and population dynamics of an early Pliocene vertebrate fauna, Knox County, Nebraska. Contributions to geology. University of Wyoming: Laramie.

Vrba ES. 1980. The significance of bovid remains as indicators of environment and predation patterns. In: Behrensmeyer AK, and Hill AP, editors. Fossils in the making: vertebrate taphonomy and paleoecology. Chicago: University of Chicago Press. p 247-271.

Vrba ES. 1995. The fossil record of African antelopes (Mammalia, Bovidae) in relation to human evolution and paleoclimate. In: Vrba E, Denton G, Patridge T, and Burcke L, editors. Paleoclimate and evolution with emphasis on human origins New Haven, CT: Yale University Press.

Vrba ES. 2006. A possible ancestor of the living waterbuck and lechwes: Kobus basilcookei sp. nov.(Reduncini, Bovidae, Artiodactyla) from the Early Pliocene of the Middle Awash, Ethiopia. T Roy Soc S Afr 61(2):63-74.

Vrba ES. 2015. Role of environmental stimuli in hominid origins. In: Henke W, and Tattersall I, editors. Handbook of paleoanthropology. Berlin, Heidelberg: Springer. p 1837-1886.

Ward C, Leakey M, and Walker A. 1999. The new hominid species Australopithecus anamensis. Evol Anthropol 7(6):197-205.

Ward C, Manthi F, and Plavcan J. 2013. New fossils of Australopithecus anamensis from Kanapoi, West Turkana, Kenya (2003–2008). J Hum Evol 65(5):501-524.

Ward CV, Leakey MG, and Walker A. 2001. Morphology of Australopithecus anamensis from Kanapoi and Allia Bay, Kenya. J Hum Evol 41(4):255-368.

Werdelin L, and Lewis ME. 2018. A contextual review of the Carnivora of Kanapoi. J Hum Evol.

Werdelin L, and Manthi FK. 2012. Carnivora from the Kanapoi hominin site, northern Kenya. J Afr Earth Sci 64:1-8.

169

Wheeler PE. 1991. The thermoregulatory advantages of hominid bipedalism in open equatorial environments: the contribution of increased convective heat loss and cutaneous evaporative cooling. J Hum Evol 21(2):107-115.

White TD. 1992. Prehistoric cannibalism at Mancos 5MTUMR-2346. Princeton: Princeton University Press.

White TD. 1995. African omnivores: global climatic change and Plio-Pleistocene hominids and suids. Paleoclimate and evolution, with emphasis on human origins. New Haven: Yale University Press. p 369-384.

White TD, Ambrose SH, Suwa G, Su DF, DeGusta D, Bernor RL, Boisserie J-R, Brunet M, Delson E, and Frost S. 2009a. Macrovertebrate paleontology and the Pliocene habitat of Ardipithecus ramidus. Science 326(5949):67-93.

White TD, Asfaw B, Beyene Y, Haile-Selassie Y, Lovejoy CO, Suwa G, and WoldeGabriel G. 2009b. Ardipithecus ramidus and the paleobiology of early hominids. Science 326(5949):64-86.

White TD, and Harris JM. 1977. Suid evolution and correlation of African hominid localities. Science 198(4312):13-21.

White TD, Suwa G, and Asfaw B. 1995. Australopithecus ramidus, a new species of early hominid from Aramis, Ethiopia. Nature 375(6526):88.

White TD, WoldeGabriel G, Asfaw B, Ambrose S, Beyene Y, Bernor RL, Boisserie J-R, Currie B, Gilbert H, and Haile-Selassie Y. 2006. Asa Issie, Aramis and the origin of Australopithecus. Nature 440(7086):883-889.

Williams SH, and Kay RF. 2001. A comparative test of adaptive explanations for hypsodonty in ungulates and rodents. J Mammal Evol 8(3):207-229.

Winkler A. 1998. New small mammal discoveries from the early Pliocene at Kanapoi, West Turkana, Kenya. J Vert Paleontol 18(3 Suppl.):87.

Wong SN, Saj TL, and Sicotte P. 2006. Comparison of habitat quality and diet of Colobus vellerosus in forest fragments in Ghana. Primates 47(4):365-373.

170

Wood B, and Harrison T. 2011. The evolutionary context of the first hominins. Nature 470(7334):347-352.

Wood B, and Leakey M. 2011. The Omo-Turkana Basin fossil hominins and their contribution to our understanding of human evolution in Africa. Evol Anthropol 20(6):264-292.

Wynn JG. 2000. Paleosols, stable carbon isotopes, and paleoenvironmental interpretation of Kanapoi, Northern Kenya. J Hum Evol 39(4):411-432.

Wynn JG. 2004. Influence of Plio-Pleistocene aridification on human evolution: Evidence from paleosols of the Turkana Basin, Kenya. Am J Phys Anthropol 123(2):106-118.

Yakir D. 1998. Oxygen-18 of leaf water: a crossroad for plant-associated isotopic signals. In: Griffiths H, editor. Stable isotopes: integration of biological, ecological and geochemical processes. Oxford: BIOS Scientific Publishers. p 147-168.

171