Isotopic Analysis and Mobility Mapping of Mammuthus Columbi from the Mammoth Site in South Dakota

Home , Pygmy mammoth, Steppe mammoth, The Mammoth Site

East Tennessee State University Digital Commons @ East Tennessee State University

Electronic Theses and Dissertations Student Works

8-2021

Isotopic Analysis and Mobility Mapping of Mammuthus columbi from the Mammoth Site in South Dakota

Matthew Harrington East Tennessee State University

Follow this and additional works at: https://dc.etsu.edu/etd

Part of the Biogeochemistry Commons, and the Paleontology Commons

Recommended Citation Harrington, Matthew, "Isotopic Analysis and Mobility Mapping of Mammuthus columbi from the Mammoth Site in South Dakota" (2021). Electronic Theses and Dissertations. Paper 3932. https://dc.etsu.edu/etd/3932

This Thesis - unrestricted is brought to you for free and open access by the Student Works at Digital Commons @ East Tennessee State University. It has been accepted for inclusion in Electronic Theses and Dissertations by an authorized administrator of Digital Commons @ East Tennessee State University. For more information, please contact [email protected]. Isotopic Analysis and Mobility Mapping of Mammuthus columbi from the Mammoth Site in

South Dakota

______

A thesis

presented to

the faculty of the Department of Geosciences

East Tennessee State University

In partial fulfillment

of the requirements for the degree

Master of Science in Geosciences, Paleontology

______

Matthew Harrington

August 2021

______

Chris Widga, Chair

Andrew Joyner

Joshua X. Samuels

Blaine W. Schubert

Keywords: Mammoth Site, Pleistocene, isotope, mobility, paleoecology

ABSTRACT

Isotopic Analysis and Mobility Mapping of Mammuthus columbi from the Mammoth Site in

South Dakota

Matthew Harrington

The Mammoth Site in Hot Springs, South Dakota preserves a unique death assemblage of sub- adult and adult male Columbian mammoths (Mammuthus columbi). Extensive work on the site has led to a detailed understanding of the taphonomy of the assemblage; yet the life histories and ecology of these mammoths remain relatively unknown. Tooth enamel from four Mammoth Site mammoth individuals were bulk sampled with one of the individuals (MSL 742) also serially micro-sampled for 훿13C, 훿18O, and 87Sr/86Sr. Isotopic results indicate that MSL 742 remained within the southern and western Black Hills year-round with no consistent migration patterns.

훿13C and 훿18O values contain minimal fluctuations, suggesting drinking water and forage was sourced from the local hot springs and surrounding landscape. This study suggests the high level of sloped landscapes in the region may have resulted in a “bull”-only region, explaining the absence of females and juvenile mammoths at the site.

ACKNOWLEDGEMENTS

I would like to acknowledge the Geological Society of America (GSA) for providing a grant that made this research possible. I thank my committee chair, Dr. Chris Widga, as well as my other committee members Dr. Andrew Joyner, Dr. Joshua X. Samuels, and Dr. Blaine

Schubert for their input and assistance with this project. I want to thank the Mammoth Site, including Dr. Jim Mead and Olga Potapova, for letting me access collections and provide helpful insights into the paleontology of the site. I also want to thank Dr. Al Wanamaker and Dr. Doug

Walker for their collaborative assistance analyzing samples in their respective isotopic laboratories.

TABLE OF CONTENTS

ABSTRACT ...... 2 ACKNOWLEDGEMENTS ...... 4 LIST OF TABLES ...... 7 LIST OF FIGURES ...... 8 CHAPTER 1. INTRODUCTION ...... 10 Introduction ...... 10 Stable Carbon Isotopes ...... 10 Stable Oxygen Isotopes ...... 12 Strontium Isotopes ...... 15 Mammoths...... 18 Isotope Studies in Proboscideans ...... 20 The Mammoth Site ...... 23 CHAPTER 2. A SURFACE 87SR/86SR ISOSCAPE MODEL FOR THE BLACK HILLS AND SURROUNDING REGION ...... 27 Introduction ...... 27 Methods ...... 29 Results ...... 32 North American Interior Model ...... 32 Rockies/Great Plains Model ...... 42 Discussion ...... 51 Model Comparison ...... 51 Regions with High 87Sr/86Sr Values ...... 53 Conclusion...... 54 CHAPTER 3. ISOTOPIC ANALYSIS AND MOBILITY MAPPING OF MAMMUTHUS COLUMBI FROM THE MAMMOTH SITE ...... 55 Introduction ...... 55 Materials and Methods ...... 60 Results ...... 66 Bulk Samples ...... 66 MSL 742 ...... 69

Discussion ...... 74 Influence of Hot Springs on Local Isotopic Variability ...... 74 MSL 742 Dietary Reconstruction ...... 77 Geographical Influences ...... 80 Implications for Mammoth Paleoecology ...... 87 Conclusion...... 89 REFERENCES ...... 91 APPENDICES ...... 110 Appendix A: MSL 742 Mobility Maps ...... 110 Appendix B: Geologic Age Table ...... 121 VITA ...... 139

LIST OF TABLES

Table 1. Bulk Isotopic Values from MSL 734, 742, 749, and 3068 ...... 68

Table 2. Micromill Isotopic Values from MSL 742 ...... 70

LIST OF FIGURES

Figure 1. MIS 6 maximum glaciation extent ...... 24

Figure 2. Primary 87Sr/86Sr isoscape of the North American Interior centered on the Black

Hills ...... 33

Figure 3. Primary model’s training features with prediction error ...... 34

Figure 4. Bottom 5th percentile of predicted 87Sr/86Sr values for the primary model ...... 35

Figure 5. Upper 95th percentile of predicted 87Sr/86Sr values for the primary model ...... 36

Figure 6. 90% confidence prediction interval of training features in the primary model ...... 37

Figure 7. Distribution of importance scores for each explanatory raster across all

validation runs in the primary model ...... 39

Figure 8. Distribution of importance scores with best iteration shown in the primary model .....40

Figure 9. R2 value distribution across all validation iterations in the primary model ...... 41

Figure 10. 87Sr/86Sr isoscape of the North American West centered on the Black Hills ...... 42

Figure 11. Western model’s training features with prediction error ...... 43

Figure 12. Bottom 5th percentile of predicted 87Sr/86Sr values for the western model ...... 44

Figure 13. Upper 95th percentile of predicted 87Sr/86Sr values for the primary model ...... 45

Figure 14. 90% confidence prediction interval of training features in the western model ...... 46

Figure 15. Distribution of importance scores for each explanatory raster across all

validation runs ...... 48

Figure 16. Distribution of importance scores with best iteration shown in the western

model ...... 49

Figure 17. R2 value distribution across all validation iterations in the western model ...... 50

Figure 18. Columbian mammoth molars used in the study. A – occlusal surface B – anterior

view C – lateral view ...... 61

Figure 19. Micromilling schematic for MSL742 ...... 63

Figure 20. Stable isotope results for MSL 742 across the sample span ...... 70

Figure 21. Results of 훿13C vs 훿18O in MSL 742 ...... 71

Figure 22. Results of 훿13C vs 훿18O in MSL 742 with 훿13C values shifted backwards one

sample ...... 71

Figure 23. MSL 742 total potential mobility across all 20 samples ...... 72

Figure 24. Most common areas of MSL 742’s potential mobility across all 20 samples ...... 73

Figure 25. Most common areas of MSL 742’s potential mobility across all 20 samples,

zoomed into the Southern Black Hills ...... 74

Figure 26. Slopes within the Black Hills at or above 5° ...... 83

Figure 27. Slopes within the Black Hills at or above 10° ...... 84

Figure 28. Slopes within the Black Hills at or above 15° ...... 85

Figure 29. The total mobility of MSL 742 (displayed from most to least common across all

20 samples) alongside all slopes at or greater than 5° within the Black Hills ...... 86

Figure 30. The total mobility of MSL 742 (displayed from most to least common across all

20 samples) alongside all slopes at or greater than 10° within the Black Hills ...... 87

CHAPTER 1. INTRODUCTION

Introduction

The Mammoth Site in Hot Springs, South Dakota was a natural trap for mammoths during the Illinoian glaciation. With over 60 complete mammoths excavated, the Mammoth Site provides a unique insight into mammoth paleoecology in the Black Hills and Great Plains. The mammoths at the Mammoth Site are all sub-adult to adult males, with no females or juveniles.

Reconstruction of individual life-histories is a strong method for better understanding similar ecological phenomena. To gain more in-depth analyses into the life-history of the mammoths, isotopic samples were extracted from M3 molars. 훿13C and 훿18O isotopes were extracted to determine dietary patterns, drinking water sourcing, and climatic trends. 87Sr/86Sr was extracted to map potential mobility patterns over the sample period. Each isotopic system was analyzed was interpreted both independently and in conjunction with the other systems to best determine the life-histories of each mammoth.

Stable Carbon Isotopes

While dentition and bone morphology can provide insight to dietary specializations and general feeding types, reconstructing diets at ecological timescales requires additional analytical tools such as enamel wear or stable isotope analysis (Davis and Pineda-Munoz 2016). Analysis of structural carbonate in biogenic apatite is highly reliable in tooth enamel due to enamel’s resistance to diagenesis (Quade et al. 1992; Wang and Cerling 1994; Koch et al. 1997), and tooth enamel has been shown to accurately reflect the diets of ancient organisms (Thorpe and Van der

Merwe 1987).

The usefulness of 훿13C derives from its distinct isotopic signatures in C3 and C4 plants

(Teeri and Stowe 1976). The majority of plants on Earth utilize C3 photosynthesis, including most trees, shrubs, and cool-season grasses. C4 plants are largely monocotyledons, and rarely, dicotyledons (Bender 1971). In warm or arid regions where potential evapotranspiration outpaces precipitation input, C4 plants are primarily composed of warm-season grasses (Cerling and

Ehleringer 2000).

Unlike the C3-photosynthetic pathway (Calvin cycle), C4 photosynthesis (Hatch-Slack

13 cycle) traps CO2 in the plant’s bundle sheath cells, enriching the ratio of 훿 C relative to C3 plants (Cerling and Ehleringer 2000). C3 훿13C values typically range from -35‰ to -20‰ with a median value of -27‰ (Dawson et al. 2002; Marshall et al. 2008). C4 훿13C values tend to cluster around -14‰ with a typical range of -15‰ to -11‰ (Dawson et al. 2002; Marshall et al. 2008;).

These distinct values allow for simple classification of consumed vegetation in the diet of herbivores.

δ13C values from vegetation become enriched upon consumption by herbivores and require correction for fractionation that occurs within different body tissues. To correct this

13 fractionation, Cerling and Harris (1999) calculated δ Cdiet using an enamel-diet fractionation factor α where:

α = 1.0141 and

13 13 α = (δ Csc + 1000) / (δ Cdiet + 1000).

The enrichment of 13C from diet to enamel of large herbivorous mammals is 14.1 ± 0.5‰

(Cerling and Harris 1999).

North American mammoths typically consumed grasses, sedges, or any available forage

(Cerling et al. 1999; Metcalfe et al. 2013; Lister 2014). In modern elephants, diet is often a blend of individual preference for specific plants, seasonality, and general availability (Field 1971;

Paley and Kerle, 1998; De Boer et al. 2000). In Kenya, elephants consumed increased C4 material during the biannual wet seasons, with a C3 or mixed C3 diet in drier seasons (Uno et al.

2020). This study is the first to report on mammoth 훿13C values from the Mammoth Site.

Stable Oxygen Isotopes

The use of stable oxygen isotopes (훿18O) has been essential for both large-scale climate reconstructions and more refined life-history climatic reconstructions (Grootes and Stuiver 1997;

Fox et al. 2007; Weißbach et al. 2016). The dominant oxygen isotope, 16O, is measured against the less common isotope, 18O, and compared to a modern standard to generate a 훿18O ratio value

(Higgins 2018). The most frequent application of this ratio is in the construction of climatic fluctuations and patterns (Dworkin et al. 2005). During glaciations, heavier 18O isotopes concentrate in oceans, as they precipitate out faster than 16O. This in turn leaves ice sheets and inland precipitation depleted in 18O (Grootes et al. 1993). Likewise, the further inland from the coast, the more depleted in 18O water becomes.

훿18O can be used to not only infer the climate but can also be used to track behavioral differences in individual water use and variability in these patterns through time (Metcalfe and

Longstaffe 2014). The oxygen isotopic composition of biogenic apatite in mammalian teeth is related to the internal body water from which it forms (Bryant et al. 1996). In large-bodied mammals such as proboscideans, the 훿18O signature of internal body water is strongly related to the 훿18O of surface water, which is often driven by local climate (Bryant et al. 1996). Surface 12 water from snow and rain is the primary source of drinking water for most animals. This source is sensitive to annual temperature shifts and to the amount of local precipitation, with higher

훿18O values in summer and lower 훿18O values in winter (Sponheimer and Lee-Thorp 1999;

Kendall and Coplen 2001). In contrast, larger bodies of surface water often show less seasonal variability. For instance, Lake Erie typically experiences less than a 2‰ shift in 훿18O values across the year whereas local (and much smaller) marshes in the region experienced up to 8‰

(Huddart et al. 1999). The limited seasonal 훿18O fluctuations seen within larger bodies of surface water are in part due to the averaging of seasonal inputs over time (Huddart et al. 1999). As meteoric water flows into these reservoirs, their 훿18O values contribute to a long-term average of the existing pool. In seasonally sensitive water sources fed by meteoric water, a larger proportion of 훿18O stems from precipitation events (Huddart et al. 1999).

Thus, 훿18O reflects a mixture of climate conditions at multiple scales, but also individual patterns in water use that may reflect mobility at low-resolution (Metcalfe et al. 2013). For species that lived in glacially dominated landscapes, glacial meltwater 훿18O values are lower compared to meteoric 훿18O values (Sima et al. 2006).

Reconstructing the 훿18O of ingested water also requires an adjustment for fractionation.

When 훿18O values are derived from structural carbonate, a two-part transformation is required to convert the acquired value into a phosphate value and then into a drinking water value.

Following Iacumin (1996), enamel carbonate values were first converted to phosphate:

18 18 훿 Op = 0.98* 훿 Oc – 8.5

Work by Daux et al. (2008) resulted in a similar conversion rate to account for these discrepancies. To convert the phosphate values into drinking water values, the formula

18 18 훿 Odw = (훿 Op – 23.3)/0.94

was used, following Ayliffe et al. (1992). It should be noted that converting bioapatite

18 18 18 18 훿 Oc to 훿 Op should be done with caution, as the 훿 Oc - 훿 Op relationship is often variable

18 18 (Pellegrini et al. 2011). Pellegrini et al. (2011) noted inconsistencies between 훿 Oc and 훿 Op not only between species, but within the same species and even the same tooth. Variations occurred both at random and at different cross-sections of teeth used in the study (Pellegrini et al. 2011).

18 18 However, for general purposes, the conversion between 훿 Oc to 훿 Op is still applicable for analysis.

Previous work on modern elephants and fossil elephants observed that 훿18O values accurately reflect ingested water (Longinelli 1984; Ayliffe et al. 1992). Isotopic analyses of modern elephants by Uno et al. (2020a) suggested positive spikes in 훿18O spikes are related to heavy consumption of plant water during Kenya’s wet season. Thus, it is likely that analysis of

훿18O values in extinct proboscideans will provide insight into the climate experienced by these animals and seasonally resolved variability in their diets. Fisher et al. (2003) used 훿18O values in tusks from Hot Springs mammoths to estimate the season of death. Variations in 훿18O were used to determine seasons represented in sampled tusks, with results indicating risk of death was higher in fall and spring. Work by Metcalfe and Longstaffe (2014) analyzed oxygen isotopes from Mammut americanum, the American mastodon, to determine seasonality and potential mobility behavior of individuals. The use of oxygen isotopes in reconstructing life histories is key in providing supplementary dietary data, accurate climate reconstruction data, and low- resolution mobility data.

Strontium Isotopes

Radiogenic strontium (Sr; atomic number 38) isotopes, a product of radioactive decay, have been utilized as geochemical tracers for both geological and biological processes (Capo et al. 1998). With an ionic radius only slightly larger than calcium (Ca; atomic number 20) (1.32 Å for strontium and 1.18 Å for calcium), Sr2+ substitutes for Ca2+ in minerals such as plagioclase feldspar, apatite, sulfates, and carbonates (particularly calcite, dolomite, and aragonite) (Capo et al. 1998; Bentley 2006). Strontium has four naturally occurring isotopes. Three of these are non- radiogenic, with abundances of 84Sr at 0.56%, 86Sr at 9.87%, and 88Sr at 82.53% (Capo et al.

1998). The fourth, 87Sr, with an abundance of 7.04%, is radiogenic, and forms from the β-decay of 87Rb with a half-life of 4.88 x 1010 years (Beard and Johnson 2000). 90Sr (half-life of only 30 years), a radioactive isotope produced as a byproduct of fission reactions, exists in low amounts in the modern environment (Capo et al. 1998).

Rubidium (Rb; atomic number 37) is an alkali metal with an atomic radius of 1.52 Å.

With a similar atomic radius to potassium (K; atomic number 19; atomic radius of 1.38 Å),

Rb1+often substitutes for K1+ in minerals such as muscovite, biotite, alkali feldspars, illite, and some evaporites (Capo et al. 1998; Bentley 2006). There are two naturally occurring rubidium isotopes: 85Rb and 87Rb. Radioactive β- decay occurs in 87Rb resulting in stable 87Sr (Capo et al.

1998).

Upon the formation of a mineral or rock in a closed system, the amount of 84Sr, 86Sr, and

88Sr remains constant in contrast to 87Sr, which increases over time as radioactive 87Rb decays

(Capo et al. 1998). In an open system, this becomes more complex as external 87Sr or 87Rb can leach in and original 87Sr or 87Rb can leach out of the system, altering the closed-system values.

87Sr/86Sr ratios in rocks indicate various pathways in the timing of mineral formation. Earth’s

15 mantle contains a constantly depleting concentration of Rb/Sr as 87Rb slowly decays into 87Sr

(Bentley 2006). The rocks formed from this supply, often oceanic basalts, retain the low Rb/Sr ratio and reflect the novel age of the formation (Capo et al. 1998). In ancient granites that formed over 3.5 billion years ago, the high ratio of Rb/Sr upon formation created a steeper trajectory towards higher values of 87Sr/86Sr. As such, older Proterozoic rocks tend to possess higher

87Sr/86Sr values than any younger rocks (Capo et al. 1998). Rocks that initially form with lower

Rb/Sr ratios will contain lower 87Sr/86Sr values than those that formed with an initially higher

Rb/Sr ratio.

The high mass of strontium in contrast to low-mass isotopes such as carbon, oxygen, or sulfur limits the amount of fractionation that occurs in geological and biological processes (Capo et al. 1998). This results in the 87Sr/86Sr values of any sample reflecting provenance or interactions with local geology without biological processes or climatic conditions altering the values.

The terrestrial cycle of strontium begins with the weathering and erosion of the rocks, releasing strontium into soil (or occasionally aerosols). Strontium in the soil is then taken up by plants, which are consumed by animals. Strontium is re-introduced to soils through carcass decay and re-enters ocean depositional environments through fluvial processes (Capo et al. 1998).

There, strontium is deposited in new marine carbonates, starting the cycle anew (Capo et al.

1998). In some cases, small amounts of strontium reenter the atmosphere and precipitate back on land (Capo et al. 1998).

The 87Sr/86Sr value of a rock generally depends on the age class of the rock. Ancient igneous rocks, such as Precambrian granites and gneisses possess high 87Sr/86Sr ratios whereas recently formed oceanic basalts will possess low 87Sr/86Sr ratios (Bentley 2006). Sedimentary

16 rocks, however, often vary greatly. Sandstones and shales, which are primarily composed of broken-down minerals from older rocks, contain variable Rb/Sr ratios from each of the parent rocks, creating 87Sr/86Sr ratios that vary from locality to locality (Graustein 1989). Thus, any two sandstones may not share the same 87Sr/86Sr ratio; depending on the source of their parent materials, the 87Sr/86Sr ratio may be very different. Sedimentary rocks which form from the precipitation of minerals in solution such as limestone or dolomite often have a much more uniform 87Sr/86Sr ratio (Graustein 1989). Seawater 87Sr/86Sr has fluctuated from 0.7068 to 0.7091 over the last 600 million years (Burke et al. 1982). Most limestones and dolomites thus reflect the 87Sr/86Sr of seawater at their formation mixed with local variations.

The lack of isotopic fractionation from bedrock to vegetation to animal tissue allows

87Sr/86Sr values recovered from an animal to be used as tracers for a source of origin. To constrain the possibilities for origin, a strontium isoscape presenting the likely 87Sr/86Sr values across a given region must be used. Beard and Johnson (2000) first attempted to model an

87Sr/86Sr isoscape for the continental United States using the geologic age of rock units from a digitized geologic map. However, the model was limited in its predictive power, and Beard and

Johnson (2000) noted that more advanced models would need to be applied to account for small scale geological variability.

Bataille and Bowen (2012) expanded upon Beard and Johnson’s (2000) model with a lithology-specific framework that also accounted for the age of the rock. Bataille and Bowen

(2012) also created a water model to predict the 87Sr/86Sr input into streams from weathering and a model based upon water catchment for a more accurate reflection of water 87Sr/86Sr content.

Further work (Bataille et al. 2014) added additional sub-models, local variability, and uncertainty assessments. This model was limited in spatial scope to the state of Alaska and further improved

17 the accuracy for 87Sr/86Sr prediction from bedrock models. The most recent attempt to model a strontium isoscape is from Bataille et al. (2018), where a random forest regression was used with covariates to predict 87Sr/86Sr values across a defined region. This model utilized a combination of the Bataille et al. (2014) bedrock model with abiotic covariates to generate a more accurate isoscape. Though applied specifically to Europe in their research, this model has potential for application globally.

Mammoths

The order Proboscidea has a rich and complex evolutionary history reaching back to the

Paleocene (Shoshani 1998; Gheerbrant 2009) and the genus Mammuthus originated in Africa around 5 Ma during the Pliocene (Maglio 1973; Wei et al. 2010). Early forms included the

African M. subplanifrons and M. africanavus in the Pliocene. The first mammoth to enter

Europe, M. rumanus, preceded the southern mammoth, M. meridionalis (Lister and Essen 2003).

Lister and Sher (2015) proposed a model of the dispersal and evolution of the ancestral M. meridionalis into the more recent M. columbi and M. primigenius in the New World. In this model, M. trogontherii, evolving from M. meridionalis, dispersed into the New World, and evolved into M. columbi. The remaining M. trogontherii population in Eurasia evolved into M. primigenius and dispersed across all of Eurasia and into North America (Lister and Sher 2015).

Van der Valk et al. (2021) provided additional context into this scenario, using ancient DNA to refine the evolutionary pathway from southern mammoths to steppe, woolly, and Columbian mammoths.

Mammoth populations in North America are tentatively separated into four distinct species: M. columbi, M. exilis, M. jeffersonii, and M. primigenius (Pasenko and Schubert 2004;

Agenbroad 2005; Saunders et al. 2010). The Columbian mammoth, M. columbi, and the woolly mammoth, M. primigenius, are the two dominant and widespread species across most of North

America. The Columbian mammoths’ range spanned over the Southwest, across the Great

Plains, to Texas and Florida (Agenbroad 2005). The woolly mammoth was found in Alaska, along the ice sheets, in the Great Lakes region, and across the north Atlantic seaboard (Claesson et al. 2017; Widga et al. 2017a). Mammuthus exilis, the pygmy mammoth, is considered an island dwarf of the Columbian mammoth and is only found on the Channel Islands off southern

California (Agenbroad 2003). Mammuthus jeffersonii, the Jeffersonian mammoth, retains a blend of morphological features of Columbian and woolly mammoths and was primarily distributed in the upper Midwest (Pasenko and Schubert 2004; Saunders et al. 2010).

Although each species possesses certain characteristic dental morphologies that can aid in identification, there tends to be large interspecific and intraspecific variation and considerable morphological overlap (Saunders et al. 2010; Widga et al. 2017b). DNA work has shown that populations of M. columbi and M. primigenius were capable of introgressive hybridization with each other (Enk et al. 2016). Recent ancient DNA work by van der Valk et al. (2021) on eastern

Siberian mammoths suggest that the Columbian mammoth lineage can be attributed to a hybridization event between a previously unrecognized mammoth lineage with an early woolly mammoth lineage. These results corroborate the proposed dispersal of mammoths into North

America by Lister and Sher (2015) and aid in the explanation of successful woolly mammoth and Columbian mammoth introgression.

Dental eruption in mammoths (and Proboscidea) are unusual because all cheek teeth erupt in the rear of the mouth and migrate forward. Mammoths possess six cheek teeth per jaw that form and wear successively throughout their life (Kozawa et al. 2002). The molariform teeth

19 are referred to, in successive order, as dp2, dp3, dp4, m1, m2, and m3 or m1, m2, m3, m4, m5, m6 (Anders and Koenigswald 2013). Each tooth is composed of enamel-coated dentin plates held together by cementum (Haynes 1991; Ferretti 2003). Both the enamel and dentin preserve well in mammoth teeth, but with dentin more prone to diagenesis, enamel is the primary source of isotopic data in the teeth (Koch et al. 1997). Each tooth possesses a unique size and numeric range of enamel plates that form over many years of a mammoth’s life. The state of dental eruption and wear can be used to determine age at death.

Due to the longevity of mammoths, a fully formed adult mammoth tooth is a relatively long-term repository for isotopic data. Metcalfe and Longstaffe (2012) showed an average enamel extension rate of 13 to 14 mm per year between the base and top of the molar, with daily extension rates falling between 62.5µm to 32.3 µm (Dirks et al. 2012). Extension rates refer to the ameloblasts (enamel-forming cells) rate of differentiation in the tooth, impacting the speed of increase in plate height (Dirks et al. 2012). Fully formed adult Columbian mammoth teeth take

10 to 13 years to form and are retained for over a decade of an animal’s life (Uno et al. 2020b).

This slow extension rate allows isotopic data to be observed at scales rarely encountered in paleontology. Widga et al. (2020) separated paleoecological scales into three distinct classes: small (days), meso (weeks to months), and large (years to decades). Small and meso-scale paleoecological data provides insights that are more comparable to extant species (Widga et al.

2020).

Isotope Studies in Proboscideans

Isotopic analyses of proboscideans have provided insights into geographically specific diet and mobility patterns of mammoth populations. Analyses have also illuminated the impact

20 of climate on different populations across time and space. As isotopic research incorporates more biogeographic diversity, a more complete picture of mammoth ecology can be constructed.

Results from different geographic regions often confirm population-wide trends, but also highlight population-specific behaviors.

Mammoths from Florida in the late Pleistocene were primarily grazers, with specimens from the Aucilla River fauna almost exclusively consuming C4 grasses (Hoppe and Koch 2006).

In the American Southwest, mammoths consumed a higher quantity of C4 plants in the eastern portion of the region in contrast to the western portion of the region; this pattern was suggested to reflect the availability of C4 grasses (Connin et al. 1998). Southwestern mammoths peaked their C4 consumption in summer with less consumption in winter months (Metcalfe et al. 2011).

In Missouri, the prevalence of dietary C4 varied between interglacial, Mid-Wisconsin, and late- glacial mammoths, with a heavy C3 diet during the late-glacial, and a C4-heavy diet during the

Sangamon Interglacial and Mid-Wisconsin (Widga et al. 2020).

Dietary analysis of mastodons from the Great Lakes region resulted in low, but clearly recognizable seasonal 훿13C fluctuations with an average amplitude of 1.5-1‰ (Metcalfe and

Longstaffe 2014). The relationship between 훿13C and 훿18O values of the Great Lakes mastodons was strong, with clearly identifiable seasonality. The amplitude of 훿18O among individual mastodons was higher, ranging from 5.5‰ to 8.6‰, likely reflecting the sourcing of drinking water from smaller precipitation-fed puddles or streams in contrast to vastly larger Lake Erie

(Metcalfe and Longstaffe 2014).

훿18O values from a micro-sampled mastodon and mammoth from the Aucilla River fauna in Florida exhibited low amplitudes of only 1.8‰ and 1.2‰ (Hoppe and Koch 2006). Seasonal variability in modern meteoric water from this region is 4.5‰; thus, the seasonal amplitude in 21 these proboscideans was lower than expected. Hoppe and Koch (2006) suggested either there was a larger reservoir effect on ingested water, their microsampling method encompassed multiple seasons, or that the lag time between the consumption of 훿18O and implementation of

훿18O in proboscideans may have dampened the seasonal signal. 훿18O values from Missouri mammoths displayed individual amplitudes of around 4‰, with one individual possessing an amplitude of 7‰ (Widga et al. 2020).

Mammoths from the Aucilla River fauna in Florida were also bulk-sampled and micro- sampled for 87Sr/86Sr, revealing consistently low 87Sr/86Sr values among the bulk samples and uniformly low 87Sr/86Sr values within the micro-sampled mammoth (Hoppe and Koch 2006).

The bulk samples likely reflect multi-year averages of dietary input. The samples did not contain evidence of the higher 87Sr/86Sr in the southern Appalachians, indicating that many mammoths likely remained in central or southern Florida ecosystems year-round (Hoppe and Koch 2006).

However, relatively uniform 87Sr/86Sr values across Florida’s coastal plains suggest that some individuals could have traveled up to 400 km within the state (Hoppe and Koch 2006). The micro-sampled individual contained such a small 87Sr/86Sr range, it likely never traveled long distances (Hoppe and Koch 2006; Hoppe and Koch 2007).

Analysis of 87Sr/86Sr from the Waco Mammoth National Monument did not uncover any systematic migration patterns in the mammoths (Esker et al. 2019). Mammoths from Missouri also failed to show any consistent, seasonal migration patterns. Widga et al. (2020) noted that some individuals engaged in minimal travel, remaining in the western Ozarks, while another moved 250 km at some point between the cessation of growth of its M3 and its ultimate death.

Evidence across the continental United States indicates mammoths did not engage in consistent seasonal movements but were capable of one-way or sporadic long-distance movements. With

22 no clear-cut migration patterns amongst North American mammoth populations, mammoth mobility is more consistent with movement related to ecological or individual life-history events such as male dispersal from matriarchal herds, social dynamics, or climatic events.

The Mammoth Site

The Mammoth Site in Hot Springs, South Dakota, discovered in 1974, has become a unique and world-renowned site for late Pleistocene mammoth paleontology (Agenbroad 1994a).

Accidentally uncovered by a housing development construction crew, it was quickly realized how important and bone-heavy the site was. Excavation and analysis of the site has continued since its initial discovery, creating a well-documented and well-described bonebed.

The site lies as a locally topographic high point relative to the surrounding area.

Evaporites in the Minnelusa Formation resulted in various collapse events that formed breccia pipes in the region (Laury 1980). One of these breccia pipes would become the location for the

Mammoth Site. Various breakages and collapse events in the breccia pipe created a sinkhole and regional artesian springs provided an inflow of water to the sinkhole (Laury 1980). The duration of the hot springs activity resulted in three distinct sedimentation phases (Laury 1980). Each phase represents variation in flow, material, and depth. Initially, a radiocarbon date on bone apatite provided a date of 25,000 to 26,000 years BP for the mammoth assemblage (Agenbroad

1994b). More recently, it is recognized that bone apatite dating is often unreliable and, at best, can only represent a minimum age. A uranium-thorium date acquired by Curtis McKinney dated the site to 128,966 years BP and was initially deemed incorrect (Agenbroad 1994b). Recent efforts to date the site have utilized Optically Stimulated Luminescence (OSL) dating and have

23 determined an estimate of roughly 140,000 years BP at the top of the pit and 190,000 at the base of what has been excavated (Mahan et al. 2016; Jim Mead 2021 personal communication).

With the more accurate OSL age, this site dates to MIS-6 (Illinoian Glaciation), the penultimate glacial period in North America. During this stage, glaciers reached as far west as central South Dakota, and as far south as southern Illinois (Stiff and Hansel 2004). Peak glaciation during MIS 6 resulted in the Laurentide ice sheet reaching as far west as easter

South/North Dakota, including northern North Dakota, and the Cordilleran ice sheet reaching as far east as central Wyoming (Figure 1) (Batchelor et al. 2019). For at least a stretch of time, the ice-free corridor to Beringia would have been open between the two ice sheets (Batchelor et al.

2019).

Figure 1. MIS 6 maximum glaciation extent (Batchelor et al. 2019)

Pollen from the site is composed of primarily grasses and sedges (61.1%) with only a small fraction of arboreal pollen (7.1%) (Mead et al. 1994). From pollen alone, it has been assumed trees were a minor presence on a relatively open grassland environment with Mead et al. (1994) classifying the landscape as an arid shrub steppe. Analyses of Mammoth Site mollusks also support an arid shrub steppe environment (Mead et al. 1994). Although these interpretations were initially applied to a Wisconsin-aged site, the Mammoth Site’s actual placement in MIS 6

(the previous glaciation) allows the interpretations to remain valid.

Fossil mollusk analyses from the Nelson-Wittenberg Site, located nearby in the southern

Black Hills, identified a diverse molluscan fauna that indicated a more diverse vegetational setting (Jass et al. 2002). The diversity in molluscan remains pointed to the presence of a mosaic of mesic micro-habitats within the greater shrub-steppe of the region (Jass et al. 2002). Localized regions of wet meadows and woodlands likely existed in and around the springs of the southern

Black Hills, reflecting a diverse landscape in contrast to a purely homogenous shrub steppe (Jass et al. 2002). The site was dated to 37,900 ± 2900 yr. BP from a thermoluminescence date on sediments associated with the site (Jass et al. 2002). The Nelson-Wittenberg Site thus falls within the Wisconsin glaciation, in a different glaciation than when the Mammoth Site was actively trapping mammoths. However, similar climatic trends in both glaciations can allow a conservative application of Jass et al. (2002)’s interpretations to the southern Black Hills during

MIS 6. The presence of an active hot spring in the Mammoth Site for a large stretch of time indicates springs were able to exist during MIS 6. It is assumed other localized springs likely occurred throughout the Southern Black Hills during MIS 6 as they did during the Wisconsin glaciation and today.

Over 60 mammoths (primarily M. columbi, but also including two M. primigenius specimens) have been recovered from the Mammoth Site with additional remains still under excavation (Agenbroad 1994c). These skeletons vary in degree of completeness and represent sub-adult to adult male individuals with no female or juvenile specimens (Lister and Agenbroad

1994). Preference for young males has been assumed to be analogous to the behavior of modern male elephants that leave the herd. The density, quantity, and articulation of subadult and adult male mammoth material at the site creates a unique locality preserving an interesting natural bias.

Isotopic tusk work by Fisher et al. (2003) looked at the rate of growth in tusks and the season of death using high-resolution 훿18O analysis. Results indicated that the primary seasons of death were spring and fall and that mammoths may have maintained some degree of site fidelity throughout multiple years. Interpretations of this study were made when 26,000 years BP was still the accepted age date and thus may require reinterpretation.

While mammoths are the dominant fauna of the site, other Pleistocene megafauna, some of which include the short-faced bear (Arctodus simus), pronghorn (Antilocapra americana), and camel (Camelops hesternus), have also been recovered (Agenbroad 1994a; Czaplewski and

Mead 1994). In addition to megafaunal species, a wide variety of small mammals have been uncovered, including (but not limited to) the eastern mole (Scalopus aquaticus), white-tailed prairie dog (Cynomys sp.), and mink (Mustela vison) (Agenbroad 1994a; Czaplewski and Mead

1994).

CHAPTER 2. A SURFACE 87SR/86SR ISOSCAPE MODEL FOR THE BLACK HILLS AND

SURROUNDING REGION

Introduction

Radiogenic strontium (Sr; atomic number 38) isotopes, a product of radioactive decay, have been utilized as geochemical tracers for both geological and biological processes (Capo et al. 1998). 87Sr, one of four naturally occurring strontium isotopes, and the only one that is radiogenic, forms from the β-decay of 87Rb with a half-life of 4.88 x 1010 years (Bentley 2006).

The decay of 87Rb into 87Sr results in distinctive 87Sr/86Sr values within minerals and rock depending on the time of formation. In the mantle, the ratio of Rb to Sr remains constant in a relatively homogenous mixture (Capo et al. 1998). As oceanic basalts form from mid-ocean ridges, their Rb/Sr ratios reflect the low values of their source, the mantle (Bentley 2006). As such, these newer basalts contain a low 87Sr/86Sr value.

In contrast, ancient Precambrian granites and gneisses formed billions of years ago and mineralized with an initially high ratio of Rb/Sr (Capo et al. 1998). The high initial Rb/Sr ratio coupled with billions of years to decay resulted in these rocks containing high 87Sr/86Sr ratios

(Capo et al. 1998). The primary drivers of 87Sr/86Sr values within a rock are time and the initial

Rb/Sr ratio (Bentley 2006). The 87Sr/86Sr value of a rock can be used to determine both of these inputs.

Unlike low-mass isotopes of carbon, oxygen, or sulfur, the high mass of strontium limits the levels of fractionation that typically occurs in geological and biological processes (Capo et al.

1998). The terrestrial cycle of strontium begins with the weathering and erosion of the rocks, releasing strontium into soil (or occasionally aerosols). Strontium in the soil is then cycled

27 through plants, which are then consumed by animals, which then cycle back into the soil.

Eventually, strontium enters the ocean, often through fluvial processes. The 87Sr/86Sr values of animal remains thus reflect the local geology from which the strontium was sourced, allowing potential determination of provenance without biological processes or climatic conditions altering the values.

To source the potential geologic source of 87Sr/86Sr from a biological sample, an isoscape must be created. An isoscape is a continuous map that displays the predicted isotopic value across space (West et al. 2010). Isoscapes can be created by algorithmic models driven by select data, as sampling and isotopically analyzing every square mile of a continent is unfeasible (West et al. 2010).

An early attempt to model an 87Sr/86Sr isoscape by Beard and Johnson (2000) utilized digitized, continent-scale, bedrock geology maps as the sole predictor for the distribution of

87Sr/86Sr values in surficial sediments. Although a good start, it is now recognized that the surface expression of 87Sr/86Sr is driven by more complex factors.

Improving upon Beard and Johnson’s (2000) model, Bataille and Bowen (2012) developed a lithology-specific framework that connected age to rock type. Models predicting water catchment and weathering 87Sr/86Sr into streams were also created to better predict

87Sr/86Sr within water (Bataille and Bowen 2012). Further improving upon their 2012 model,

Bataille et al. (2014) added additional sub-models to better predict distinct rock types and provide uncertainty assessments more accurately.

Bataille et al. (2018) determined that random forest models outperformed other regression and interpolation methods for predicting 87Sr/86Sr isoscapes. Their model utilized

28 abiotic covariate rasters alongside bedrock models and real-world 87Sr/86Sr values to generate a strontium isoscape for western Europe.

Breiman (2001) described random forests as a collection of tree predictors where each tree samples random vectors independently with equal distribution across the forest. As the number of trees grow, the forest converges to the most likely outcome. This study looks to use

ESRI’s adaptation of Breiman’s random forest algorithm to generate an isoscape of the North

American Interior.

Methods

87Sr/86Sr values gathered across the central United States by Widga et al. (2017c) from vegetation and sediment were used as training points for a forest-based classification and regression model based upon Breiman’s random forest algorithm within ArcGIS Pro (Breiman

2001). Each 87Sr/86Sr sample gathered by Widga et al. (2017c) was targeted to reflect as much diversity within the local geology, creating a diverse dataset of 87Sr/86Sr values that can aid in prediction of unsampled regions.

The spatial scope of this isoscape encompasses large parts of the interior of North

America, from 38° – 50° latitude to 86° – 108° longitude (including the states of Colorado,

Illinois, Indiana, Iowa, Kansas, Michigan, Minnesota, Missouri, Montana, Nebraska, North

Dakota, South Dakota, Wisconsin, and Wyoming). This region was defined based on the basis of one or two conditions: 1) the state contains 87Sr/86Sr data gathered by Widga et al. (2017c) and/or

2) the state is a potential area where the Hot Springs mammoths may have roamed.

To reinforce the random forest model, covariate explanatory rasters were provided as inputs. SoilGrids, a digital soil mapping system based upon soil profile data (WoSIS) powered by ISRIC (International Soil and Reference Information Centre), was used to gather 6 covariates: clay (weight %), silt (weight %), sand (weight %), soil organic carbon (weight %), pH of water

(soil), and bulk density (kg m-3) (Hengl et al. 2017). All 6 downloaded layers represented a depth of 0 – 5 cm.

SRTM 90m DEM (digital elevation model) data were accessed from the CGIAR-CSI

GeoPortal (Jarvis et al. 2008). A global sedimentary thickness covariate was acquired from

NASA’s EarthData portal (Pelletier et al. 2016). Mean annual precipitation (mm yr-1) was acquired from WorldClim Version 2 at 1 km resolution (Fick and Hijmans 2017). Global potential evapo-transpiration and global aridity rasters were gathered from the work of Zomer et al. (2008).

To generate a covariate raster reflecting geologic age, a USGS conterminous State

Geologic Map Compilation (SGMC) map of the continental United States was downloaded from the USGS online spatial data hub (Horton 2017). The total map, a mosaic of the lower 48 state’s geologic maps, lacked cohesion among rock unit ages, rock types, and mapping methods. States such as South Dakota included Quaternary deposits within their state whereas states such as

Illinois did not. To solve the issue of inconsistent ages, following the methods of Bataille and

Bowen (2012), the listed age or age range for each individual unit from the SGMC was assessed, and the maximum age was assigned (Appendix B). For example, if a unit was listed as Jurassic to Cretaceous, a unit age of 201.3 million years was assigned. This creates overestimations instead of underestimations within the ages of non-descript geologic units. All age unit

30 descriptors that were assigned ages were pulled directly from the USGS SGMC unit descriptions within ArcGIS Pro (Horton 2017).

The forest-based classification and regression (spatial statistics) tool was used within

ArcGIS Pro. The tool is an adaptation of Leo Breiman's random forest algorithm (Breiman

2001). The tool uses an ensemble of decision trees to create predictive models. 87Sr/86Sr values gathered by Widga et al. (2017c) were used as training features and the aforementioned 12 covariates were used as explanatory rasters.

1000 trees were created for the model with a default minimum leaf size of 5. The maximum tree depth was data-driven to avoid potential overfitting. 100% of the training features were available to each decision tree, with each tree randomly sampling two thirds of the training feature data. The number of explanatory variables used to create each decision tree was set at 4.

Overly increasing the number of explanatory variables used in each decision tree can lead to overfitting of the model. A common method for setting this number is to use the square root of all numeric explanatory variables. With a square root of 3.46, the number was rounded up to 4.

10% of the training input features were reserved for model validation. Validations were run for 50 iterations to determine the models with the highest coefficient of determination (R2) values. The iteration with the highest R2 value was used for model prediction.

A second model with higher spatial resolution was created for the states of Colorado,

Kansas, Montana, Nebraska, North Dakota, South Dakota, and Wyoming. Only 87Sr/86Sr samples gathered within these states were used as training features within the model. This model was created to ensure training points east of South Dakota were not driving the model to inaccurate predictions.

Results

The 87Sr/86Sr isoscape model for the North American interior is shown alongside training features with prediction errors and the lower and upper 5% prediction intervals. The 87Sr/86Sr isoscape model for the North American Rockies and Great Plains is also shown with training features with prediction errors and the lower and upper 5% prediction intervals.

North American Interior Model

The 87Sr/86Sr isoscape created for the North American Interior ranged from 0.70883 to

0.72771 (Figure 2). Regions that reflected high 87Sr/86Sr values were the Rocky Mountains, the

Greater Yellowstone ecosystem, the Bighorn Mountains, the Black Hills, and a large portion of

Minnesota. Bodies of water including rivers and lakes were excluded from analysis and remain blank. Most of the training features within this model were predicted within half of a standard deviation (Figure 3). However, three training features, two within the central Black Hills and one on the eastern flank, predicted 87Sr/86Sr values with much less accuracy (standardized residual values of 10.52, 2.05, and -1.92).

Figure 2. Primary 87Sr/86Sr isoscape of the North American Interior

Figure 3. Primary model’s training features with prediction error

The lower 5th percentile of the 87Sr/86Sr isoscape ranged from 0.71058 to 0.70768 (Figure

4). The upper 95th percentile of the 87Sr/86Sr isoscape ranged from 0.70957 to 0.74833 (Figure 5).

Within the 5th percentile map, regions with lower 87Sr/86Sr values became more visible as regions with higher 87Sr/86Sr values were under-estimated. Conversely, in the 95th percentile map, regions with high 87Sr/86Sr values were overestimated, muting the values of the rest of the map.

Figure 4. Bottom 5th percentile of predicted 87Sr/86Sr values for the primary model

Figure 5. Upper 95th percentile of predicted 87Sr/86Sr values for the primary model

The prediction interval between the bottom 5th and top 95th percentiles remained narrow for the first 155 samples (sorted by predicted 87Sr/86Sr value) (Figure 6). The remaining 25 samples, all successively increasing in 87Sr/86Sr values, resulted in a wider margin for the prediction interval. The final 5 samples (highest 5 predicted values) exhibited the greatest interval range from bottom 5th and top 95th percentiles. Thus, the 90% confidence interval for each predicted 87Sr/86Sr value is more constrained for lower 87Sr/86Sr values (Figure 6).

Figure 6. 90% confidence prediction interval of training features in the primary model

Of the 12 explanatory-covariate rasters used in the model, soil organic content, geologic age, and elevation were the primary drivers (Figure 7). Importance scores for these covariates were 0.000489 (soil organic content), 0.000337 (geologic age), and 0.000229 (elevation). These importance scores were selected from the 33rd iteration of the model’s validation. Of the 50 iterations run with 10% of the training feature data, iteration 33 resulted in the best R2 score

(0.858) (Figure 8). Of the 50 iterations, the mean R2 score was 0.300, the median R2 score was

0.239, and the standard deviation was 0.253 (Figure 9).

Figure 7. Distribution of importance scores for each explanatory raster across all validation runs in the Interior model

Figure 8. Distribution of importance scores with best iteration shown in the primary model

Figure 9. R2 value distribution across all validation iterations in the primary model

Rockies/Great Plains Model

The Rockies/Great Plains Model ranged from 0.70874 to 0.72934 (Figure 10). As with the North American Interior Model, the Rocky Mountains, the Greater Yellowstone ecosystem, the Bighorn Mountains, and the Black Hills expressed high 87Sr/86Sr values. Most of the training features within this model were predicted within half a standard deviation (Figure 11). However, three training features within the central Black Hills predicted 87Sr/86Sr values with much less accuracy (standardized residual values of 7.01, 0.94, and -2.56).

Figure 10. 87Sr/86Sr isoscape of the North American Rockies and Great Plains

Figure 11. Rockies/Great Plains model’s training features with prediction error

The lower 5th percentile of the 87Sr/86Sr isoscape ranged from 0.70768 to 0.71016 (Figure

12). The upper 95th percentile of the 87Sr/86Sr isoscape ranged from 0.70951 to 0.74834 (Figure

13). Within the 5th percentile map, mountainous regions typically associated with high 87Sr/86Sr values became dampened, creating a much smaller range between the highest and lowest

87Sr/86Sr values. Conversely, in the 95th percentile map, regions with high 87Sr/86Sr values were overestimated while regions with lower 87Sr/86Sr values remained consistently low. This pushed the map to reflect two extremes with few values representing the middle.

Figure 12. Bottom 5th percentile of predicted 87Sr/86Sr values for the Rockies/Great Plains model

Figure 13. Upper 95th percentile of predicted 87Sr/86Sr values for the Rockies/Great Plains model

The prediction interval between the bottom 5th and top 95th percentiles remained narrow for the first 52 samples (sorted by predicted 87Sr/86Sr value) (Figure 14). The remaining 8 samples, all successively increasing in 87Sr/86Sr values, resulted in a wider margin for the prediction interval. The final 3 samples (highest 3 predicted values) exhibited the greatest interval range from bottom 5th and top 95th percentiles. Thus, the 90% confidence interval for each predicted 87Sr/86Sr value is more constrained for lower 87Sr/86Sr values.

Figure 14. 90% confidence prediction interval of training features in the Rockies/Great Plains model

Of the 12 explanatory-covariate rasters used in the model, soil organic content, geologic age, and elevation were the primary drivers. Importance scores for these covariates were

0.000461 (soil organic content), 0.00043 (geologic age), and 0.000134 (elevation) (Figure 15).

These importance scores were selected from the 14th iteration of the model’s validation (Figure

16). Of the 50 iterations run with 10% of the training feature data, iteration 14 resulted in the best

R2 score (0.969). Of the 50 iterations, the mean R2 score was 0.377, the median R2 score was

0.230, and the standard deviation was 0.357 (Figure 17).

Figure 15. Distribution of importance scores for each explanatory raster across all validation runs

Figure 16. Distribution of importance scores with best iteration shown in the Rockies/Great Plains model

Discussion Model Comparison

The two models, although stemming from the same training features and explanatory rasters, differed in their geographic extent. The primary driver for creating a more geographically limited, western model was the inconsistencies in state geologic maps. The presence or lack of surficial geology layers has the potential to confuse or incorrectly drive the model. In states such as Illinois, Wisconsin, or Minnesota, extensive glaciation during the late Pleistocene deposited and removed sediment and loess across the region (Forman et al. 1992).

However, the state geologic maps acquired from these three states only focus on bedrock layers beneath the Quaternary deposits. In the case of Minnesota, many of the state’s bedrock are ancient, Precambrian rocks that likely contain high 87Sr/86Sr values. Samples gathered by Widga et al. (2017c) reflect the 87Sr/86Sr ratio of the surface, which in many cases, may be tens of meters above the bedrock.

In the model, 87Sr/86Sr values are associated with the explanatory rasters at each sample locality. For Minnesota, the model associates the billion-year-old bedrock with the input 87Sr/86Sr training points. This can be seen in the North American Interior model where most of the state is modeled as high 87Sr/86Sr values (Figure 2). Although this may be accurate in regions of

Minnesota that contain exposed bedrock or limited glacial deposits, it is unlikely to accurately reflect surfaces covered by glaciogenic sediments in other parts of the state.

In contrast, South Dakota contained Quaternary deposits in their state geologic map. The eastern half of the state is dominated by these deposits. Illinoian and Wisconsin-aged glacial deposits are common in this part of the state. The western half of the state was unglaciated and surface sediments are derived from older bedrock layers (Horton 2017).

When comparing eastern and western South Dakota, there is no clear difference in

87Sr/86Sr values (Figure 2). The primary deviance is the high values within the Black Hills, an elevated region of ancient, Precambrian rock. Excluding the Black Hills, much of western South

Dakota contains rocks dating to the early/mid Cenozoic or Mesozoic. Despite the age difference from these regions in the west to the late Quaternary deposits in the east, there is not a clear difference in predicted 87Sr/86Sr values.

Rocks older than 300 million years, and particularly those greater than 1 billion years old, strongly impact the model’s prediction of 87Sr/86Sr values. While it is effective for regions such as the Black Hills where these rocks are exposed, it is driving the prediction in Minnesota to similar results with the lack of Quaternary-aged layers in the geologic map.

To determine if this false association in Minnesota was impacting prediction accuracy in the target western South Dakota region, the North American Rockies/Great Plains model was created, removing eastern states that failed to include Quaternary deposits in their state geologic maps (Figure 10).

The removal of these states also removed a large number of training features. While the loss of additional training points can impact the model’s accuracy, it was the main way to determine if eastern states were driving the primary model towards incorrect predictions. The

Rockies/Great Plains model still retained 60 training points (Figure 14).

When compared, the two models were very similar. Both models contained a similar minimum value, with a difference of only 0.00009 between the two (Figures 2 and 10). The maximum predicted value varied slightly more, with a difference of 0.00163 (Figures 2 and 10).

Despite the inclusion of eastern states in the Interior model, prediction results in the western

52 states remained consistent between the two models. It is likely the additional 120 training features included in the Interior model (excluded in the western model) provided enough baseline data to override any potentially incorrect association data from eastern states lacking surficial geological data (Figure 6). Due to the exclusion of 120 training features within the

Rockies/Great Plains model, the Interior model was used for isoscape application.

Regions with High 87Sr/86Sr Values

In both models, the predictions for higher 87Sr/86Sr values expressed the widest 90% confidence intervals (Figures 6 and 14). In regions with lower 87Sr/86Sr values, the range between the 5th and 95th percentile fell around 0.001 to 0.0015 (Figures 6 and 14). In regions with higher

87Sr/86Sr values, that interval grew considerably, reaching a maximum of 0.0388 (Figures 6 and

14).

In predictive models, it is difficult for the algorithm to predict values that are higher than the given training features. The majority of samples gathered by Widga et al. (2017c) contained low or moderate 87Sr/86Sr values. With a limited pool of high 87Sr/86Sr values in the training feature data set, any model will likely underpredict 87Sr/86Sr values in certain regions.

In the case of the Black Hills, both models struggled to predict the higher 87Sr/86Sr values found in the central Black Hills (Figures 3 and 11). In the southern Black Hills and eastern outskirts, the model was more accurate and successful. If additional 87Sr/86Sr samples had been taken in regions with high 87Sr/86Sr values, the model would likely predict these regions with more consistency.

Conclusion

An 87Sr/86Sr isoscape for the North American Interior was generated using a random forest predictive model. Geographically spaced 87Sr/86Sr samples were used as training features within the model. Twelve explanatory rasters were utilized as covariates to explain 87Sr/86Sr distribution. A second model was created under the same parameters but was restricted to seven western states.

Soil organic carbon, geologic age, and elevation exhibited the strongest importance scores for both models. The western model did not result in significant differences from the primary model. Geologic age of Precambrian rocks drove the primary model to overpredict the

87Sr/86Sr ratio in Minnesota, as Quaternary deposits were not included in the USGS geologic map of the state.

CHAPTER 3. ISOTOPIC ANALYSIS AND MOBILITY MAPPING OF MAMMUTHUS

COLUMBI FROM THE MAMMOTH SITE

Introduction

The cause of megafaunal extinctions at the end of the Pleistocene has been the cause of debates for decades (Martin 1973; Haynes 2002; Barnosky et al. 2004; Koch and Barnosky 2006;

Grayson and Meltzer 2015). The rapid disappearance of most of North America’s largest megafaunal species across such a wide and diverse geographic extent encourages a shift in focus from continental extinction to regional extinctions (Emery-Wetherell et al. 2017; Widga et al.

2017a; Broughton and Weitzel 2018). But to better interpret these extinction timelines and processes, understanding the progressive impact of humans and climate changes on keystone species is key (Grayson 2007; Pineda-Munoz et al. 2021). Individual behaviors and patterns can aid in the interpretations of broader, population-wide ecological dynamics when combined and applied together (Wittemyer et al. 2005; Fernando et al. 2008).

As a keystone species, mammoths were almost uniformly present in most ecosystems across North America (Agenbroad 2005). Populations were present in all regions of the continent not covered by ice sheets, ranging from Alaska to Mexico and California to Florida. Historically,

North American mammoths have been divided into four species: Columbian mammoths

(Mammuthus columbi), woolly mammoths (Mammuthus primigenius), Jeffersonian mammoths

(Mammuthus jeffersonii), and pygmy mammoths (Mammuthus exilis) (Agenbroad 2005; Lister

2014; Lister 2017). This understanding of population dynamics has changed in recent years.

Mitochondrial DNA work by Enk et al. (2016) suggested that Columbian mammoths and woolly mammoths were probably able to interbreed and might be better visualized as a single, morphologically variable but genetically viable metapopulation. Morphometric data of

Jeffersonian mammoths indicated they may have been the outcome of the complex introgression between woolly and Columbian mammoths (Lister 2017). Tooth morphology follows a similar pattern, with variability seemingly tracking biogeography more than specific evolutionary lineages (Widga et al. 2017a). Recent ancient DNA work by van der Valk et al. (2021) on eastern

Siberian mammoths uncovered that the Columbian mammoth lineage can be attributed to a hybridization event between a previously unrecognized mammoth lineage with an early woolly mammoth lineage. These results corroborate the proposed dispersal of mammoths into North

America by Lister and Sher (2015).

This geographic variability seen in mammoth populations across North America casts doubt on the appropriateness of using modern elephant populations as ecological and behavioral analogues for mammoths. Modern elephants, consisting of African bush elephants (Loxodonta africana), African forest elephants (Loxodonta cyclotis), and Indian elephants (Elephas maximus), are primarily restricted to warm climates, with most populations inhabiting savannas or rainforests. During the late Pleistocene of North America, mammoth habitat ranged from the arid and cold mammoth steppe of Alaska to coastal plains of Florida and to the moist, cool deserts of the Southwest (Koch et al. 1998; Agenbroad 2005; Holmgren et al. 2007; Rivals et al.

2010). Although mammoths inhabited a wider range of landscapes than elephants, their similarity in size and body plan suggests that modern elephant ecology and behavior may be comparable to mammoths.

Compared to smaller mammals, proboscideans require large amounts of food to support a population. Extant elephants consume 100 to 200 kg of food per day (McCullagh 1969). When elephants forage together in herds or larger social groups, they have a significant impact on local vegetation (Barnes 1983). As forage quality decreases, elephants are often forced to relocate to

56 new food sources or face starvation (Wato et al. 2016). In addition to available forage, elephant mobility is strongly tethered to permanent water sources and by the timing and amount of seasonal precipitation (Osborn 2004; de Knegt et al. 2011). Home range sizes in wetter environments are often smaller than those in dry environments; and precipitation levels can be a strong driver for both vegetation dynamics and changes in population density (van Aarde et al.

2008).

Understanding the drivers behind extant elephant range size and movement patterns creates the potential for ecology-based interpretations of extinct mammoth populations. Extant elephant behavior and landscape-use patterns can be used to generate behavioral hypotheses of extinct proboscideans, especially if they are formulated in terms that can be tested through paleoecological methods such as stable isotope analyses. Previous research on mammoths from

Texas (Esker et al. 2019), Missouri (Widga et al. 2020), and Illinois (Harrington et al. 2019) suggest that individual mammoths did not migrate seasonally, or even demonstrate consistent, inter-annual movements. Mammoths from these regions exhibited local movement or engaged in sub-regional, unidirectional movement over multiple years. Investigation of mammoth movement behavior from other regions of North America will aid in better understanding factors underlying mammoth mobility.

Due to their size, mammoths are often found as isolated teeth or bones. Occasionally, isolated but more complete skeletal remains are found, especially in regions with favorable depositional environments. The Mammoth Site, in Hot Springs, South Dakota, first uncovered in

1974, is a unique and world-renowned site for late Pleistocene mammoth paleontology

(Agenbroad 1994a). Over 60 individual mammoths have been recovered from the site (primarily

M. columbi, but also including two M. primigenius specimens) with additional remains still to be

57 uncovered (Agenbroad 1994c). The site is dominated by sub-adult to adult male individuals with no female or juvenile specimens (Lister and Agenbroad 1994). The density, quantity, and articulation of mammoth material at the site highlights the importance of the locality that preserves an interesting natural bias.

The site is located on a topographic high relative to the surrounding area. Evaporites in the Minnelusa Formation resulted in various collapse events that allowed the formation of breccia pipes in the region (Laury 1980). Various breakages and collapse events in the breccia pipe created a sinkhole and regional artesian springs provided an inflow of water to the sinkhole

(Agenbroad 1994b). Recent efforts to date the site using Optically Stimulated Luminescence

(OSL) and determined the top of the deposit dates to 140,000 years and the bottom to 190,000 years (Mahan et al. 2016; Jim Mead Personal Communication, 2021). This means that the paleontological assemblage preserved in this locality dates to the Illinoian Glaciation (MIS 6), the penultimate glacial period in North America.

Previous research into the isotopic ecology of mammoths has developed a strong baseline for understanding mammoth diets and landscape-use across different regions of North America.

Enamel 훿13C accurately reflects the diets of ancient organisms (Thorpe and Van der Merwe

1987). C3 and C4 plants contain distinct 훿13C isotopic signatures with C3 훿13C values typically ranging from -35‰ to -20‰ with a median value of -27‰ (Dawson et al. 2002; Marshall et al.

2008). C4 values tend to cluster around -14‰ with a typical range of -15‰ to -11‰ (Dawson et al. 2002; Marshall et al. 2008).

Mammoth diets are generally related to the availability of specific forage types. Florida

Columbian mammoths grazed heavily on C4 grasses, which the authors inferred as a reflection of local biomass (Hoppe and Koch 2006). Likewise, in the American Southwest, the relative

58 proportions of C4 grasses and C3 plants in mammoth diets reflects a regional gradient in C4 grass abundance from east to west (Connin et al. 1998). In Missouri, mammoths consumed primarily C3 plants during the recent late-glacial but consumed primarily C4 grasses during the

Sangamon Interglacial (Widga et al. 2020). Seasonal variation can also affect the productivity and accessibility of C3 vs C4 plants, with C4 abundance peaking during summer (Ode et al.

1980).

훿18O of mammoth tooth enamel has the potential to provide insights into multiple facets of mammoth life history, including seasonal climate variability, sourcing drinking water, and even spatial mobility. The oxygen isotopic composition of biogenic apatite in mammalian teeth is related to the body water from which it forms. The isotopic composition of body water itself is controlled by ingested water often in the form of drinking water (Bryant et al. 1996). In large mammals (over 1 kg) drinking water has a large impact on oxygen isotope values of animal tissues (Bryant and Froelich 1995).

Micro-sampled mammoths from Missouri displayed seasonally fluctuating 훿18O values that were interpreted to reflect seasonality (Widga et al. 2020). Likewise, mastodons (Mammut americanum) from the Great Lakes displayed strong 훿18O fluctuations; Metcalfe and Longstaffe

(2014) inferred the source of drinking water in mastodons was precipitation-fed, surface water sources rather than large permanent bodies of water. Unlike large, stable bodies of water, meteoric water found in drinking sources such as ponds or marshes is sensitive to annual temperature shift and seasonal precipitation inputs (Sponheimer and Lee-Thorp 1999).

To track individual mammoth mobility with greater precision than 훿18O, 87Sr/86Sr ratios are used. The ratio of 87Sr/86Sr within geologic units is a product of the initial Rb/Sr ratio and the age of formation (Bentley 2006). The high mass of strontium limits its fractionation within 59 biological and geological processes, allowing an 87Sr/86Sr ratio within an animal to reflect its spatial origin. Thus, 87Sr/86Sr variations within an organism’s tissues can be used to trace mobility, migration, or foraging shifts (Capo et al. 1998).

At the Waco Mammoth National Monument, 87Sr/86Sr analysis by Esker et al. (2019) did not uncover any systematic migration patterns within the matriarchal mammoth herd from the site. Similarly, Missouri mammoths failed to show consistent, seasonal movements but amongst individuals, mobility ranged from little movement to more than 250 km (Widga et al. 2020).

Mammoth mobility is likely driven by complex interactions between forage and water availability, sub-adult male dispersal from matriarchal herds, and individualistic preferences.

In this study, four upper 3rd molars from four Columbian mammoths from the Mammoth

Site were micro-sampled for 훿13C, 훿18O, and 87Sr/86Sr. Two teeth were systematically sampled via high-resolution micromilling to track diet, climatic changes, and mobility in short (4 to 6 week) intervals over the course of roughly two years. Mobility was analyzed alongside both dietary 훿13C values and 훿18O values to examine potential ecological relationships. The resulting data are compared with similar studies from Missouri, Texas, and Illinois to discuss spatio- temporal dimensions in the ecology of North American mammoths.

Materials and Methods

Four M3 molars (MSL 734, MSL 742, MSL 749, MSL 3068) from the Mammoth Site in

Hot Springs, South Dakota, were borrowed for sampling (Figure 18). Most molars from the

Mammoth Site are either embedded in the skull or on display in situ for visitors. The selected

60 molars were chosen, in part, because of their accessibility as isolated molars. All four molars were well-preserved with limited fracturing or instability.

Figure 18. Columbian mammoth molars used in the study. A – occlusal surface B – anterior view

C – lateral view

MSL 734 contained 10 preserved plates with an estimated 50% crown length completion.

Plate wear had not yet commenced on this specimen and the age of the individual is estimated to be 28 (± 2) to 30 (± 2) African Elephant Years (AEY) following Laws (1966). MSL 742

61 contained 10 preserved plates with 50% crown length completeness. The molar is about 40-50% worn and falls in the AEY age range of 36 (± 2) to 39 (± 2). MSL 749 contains 6 preserved plates with only a 25-30% crown length completeness. The tooth is 82% worn and is aged at 43

AEY. MSL 3068 contains 16 preserved plates and has retained almost all the crown length at 90-

100%. This molar is roughly 45% worn and is aged at 22 (± 2) AEY.

All four upper molars (MSL 734, 742, 749, 3068) were bulk sampled with a 2mm diameter Dremel bit along an enamel plate to gather time-averaged samples reflecting approximately one year. Although lacking in temporal precision, bulk samples provide quick overview data on one year of an animal’s life without the heavy financial and lab time costs of micro-milling. Each molar was bulk sampled twice; one sample for structural carbonate analysis and one for phosphate analysis.

Due to enamel’s resistance to diagenesis, the isotopic analysis of structural carbonate in biogenic apatite is relatively stable (Quade et al. 1992; Wang and Cerling 1994; Koch et al.

1997). To ensure enamel had not undergone diagenesis or been contaminated, half of each bulk sample was chemically treated with 200μL of 2.5% NaOCl for 24 hours and 0.1M of Acetic

Acid for 4 hours to remove any contaminants. Both the chemically treated half and the untreated half of each sample were analyzed for 훿13C and 훿18O to check for uniformity between the two.

This ensures that the micromill samples, which are too small to be chemically treated, are providing unaffected and undistorted data.

Molar MSL 742 was systematically micromilled to acquire chronologically ordered samples representing shorter amounts of time than bulk samples. A micromill with a 0.5mm diameter bit was used to collect micro-samples at 1 mm intervals, perpendicular to the growth axis of the enamel plate (Figure 19). Each sample was milled in 100 micron increments until the base of the enamel was reached (sample closest to the enamel dentin junction (EDJ)). Samples from the EDJ were chosen for isotopic analyses to minimize time averaging (Zazzo et al. 2006).

The innermost enamel layer along the EDJ is denser relative to the rest of the mature enamel and mineralizes with less temporal discrepancies (Suga 1989). The enamel powder from micromilling was collected in deionized water to decrease sample loss and to lubricate the drill bit.

Figure 19. Micromilling schematic for MSL742

Tooth formation in Columbian mammoths possess an enamel growth rate in the cervical- occlusal direction of approximately 1-2 mm per month (Dirks et al. 2012). Thus, systematically sampling at 1 mm intervals averages ~4-6-weeks of enamel growth, rather than the annual or lifetime averages provided by bulk sampling.

Samples were analyzed for 훿13C and 훿18O via a Thermo Scientific Delta V Plus mass spectrometer in continuous flow mode connected to a Gas Bench with a CombiPAL autosampler at Iowa State University’s Department of Geological and Atmospheric Sciences Stable Isotope

Lab. Reference standards NBS-18 and IAEA 603 were used to calibrate 훿13C and 훿18O values to

VPDB and were utilized for regression-based isotopic corrections (Coplen 1995; Coplen et al.

2006). The analytical uncertainty for 훿13C is ± 0.10‰ (VPDB) and 훿18O is ± 0.09‰ (VPDB).

훿13C values from vegetation become less negative when consumed by herbivores and require adjustment to account for fractionation in body tissues. 훿13C data gathered in this study

13 were adjusted for enamel-diet fractionation following Cerling and Harris (1999). δ Cdiet was calculated using an enamel-diet fractionation factor α where:

α = 1.0141 ± 0.0005 and

13 13 α = (δ Cenamel + 1000) / (δ Cdiet + 1000).

Fractionation factor α has been shown to be consistent among large herbivorous mammals (Cerling and Harris 1999). To estimate the percentage of C4 plants in the diet, a dual endpoint mixing model utilizing the distribution of 훿13C values in modern vascular plants was used (following Post 2002; Sponheimer et al. 2003). Thus:

13 %C4 = [훿 Cdiet - ( -26.5)] / [ -12.7 - ( -26.5)] * 100.

Drinking water 훿18O values can be extrapolated from enamel carbonate. 훿18O values were converted from VPDB to VSMOW using the equation:

훿18O (VSMOW) = (1.03086 * 훿18O (PDB)) + 30.86

To acquire drinking water 훿18O values, a two-step calculation was performed. Following

Iacumin (1996), enamel carbonate values were first converted to phosphate:

18 18 훿 Op = 0.98* 훿 Oc – 8.5

Secondly, following Ayliffe et al. (1992),

18 18 훿 Odw = (훿 Op – 23.3)/0.94

18 18 It should be noted that converting bioapatite 훿 Oc to 훿 Op should be done with caution,

18 18 as the 훿 Oc - 훿 Op relationship can often be variable (Pellegrini et al. 2011).

87Sr/86Sr was analyzed at the University of Kansas Isotope Geochemistry Laboratory.

Each sample was dissolved in 7.5 N HNO3 followed by eluting the Sr through ion exchange columns filled with strontium-spec resin. Using a Thermal Ionization Mass Spectrometer

(TIMS), an automated VG Sector 54, 8-collector system with a 20-sample turret, the 87Sr/86Sr ratios were then measured and adjusted to correspond to a value of 0.71250 on NBS-987. An

86Sr/88Sr value of 0.1194 was applied to correct for fractionation. Each sample has an uncertainty of +/- 0.00001.

To locate the possible regions of origin for each 87Sr/86Sr value, samples were individually mapped onto a strontium isoscape in ArcGIS Pro (see Chapter 2). Each sample’s spatial range was predicted on the isoscape within a half standard deviation above and below the samples 87Sr/86Sr value to account for weekly averaging; a +/- range of 0.00018. To refine the

65 origin of each sample, regions that were too far removed from the mammoth’s viable range, too isolated, or blocked by geographic barriers were filtered from the isoscape. The predicted spatial range for each of the 20 samples was summed to target regions that were most likely utilized by

MSL 742 across the sample span.

Results

Bulk sample 훿13C and 훿18O results for each of the molars is provided in Table 1.

Micromill sample 훿13C, 훿18O, and 87Sr/86Sr results for MSL 742 are provided in Table 2. Figures

20, 21, and 22 provide graphical comparisons of different isotopic systems. Total mobility for

MSL 742 is shown in Figure 23, with cumulative mobility shown in figures 24 and 25.

Bulk Samples

In all four molars, there was a difference between treated and untreated bulk samples.

The difference in 훿13C and 훿18O was uniform in all four molars, indicating all four specimens, each from a different individual, shared the same slight shift in values following chemical treatment. The percent difference in 훿13C was 6% to 7% across all four molars, with each treated sample enriched in 훿13C by ~0.6 (Table 1). The percent difference in 훿18O was -6% to -9% across all four molars, with each treated sample depleted in 훿18O by ~0.8 to ~1.1 (Table 1). This difference in value between untreated and treated samples is not significant enough to influence interpretation of 훿13C and 훿18O values in this study. The uniformity of the percent difference in four unique molars uncovered from different depths and locations in the Mammoth Site warrants further research.

Bulk sample 훿13C results from all four molars indicate a predominantly C3 diet, ranging from -23.62‰ to -22.66‰ with C4 plants accounting for 19% to 28% of the diet (Table 1). Bulk

18 18 훿 Osc values from all four mammoths ranged from 16.97‰ to 19.39‰ (Table 1). Bulk 훿 Odw values ranged from -16.13‰ to -13.61‰ (Table 1). The modern 훿18O isotopic composition of precipitation in South Dakota ranges from -26‰ to -22‰ in January to -6‰ to -2‰ in July, with

18 a weighted annual range of -14‰ to -10‰ (IAEA 2001). The annual average of 훿 Odw values in these specimens falls at the lower end and slightly below that of the modern annual isotopic average. Cooler climatic conditions during the Illinoian Glaciation likely account for the more negative range.

Table 1. Bulk Isotopic Values from MSL 734, 742, 749, and 3068

MSL 742

MSL 742 was micro-sampled in two parallel columns along the broad surface of the enamel plate. One column provided samples for analyses of light isotopes (훿13C, 훿18O) while the adjacent, corresponding column provided samples for 87Sr/86Sr analyses. This system allowed each micromill sample of 훿18O and 훿13C to be paired with an 87Sr/86Sr ratio, thus linking all three systems (Figure 20). The sampling increment spans roughly 2 years of the mammoth’s life.

In the 훿18O and 훿13C column, 20 samples were analyzed. Only 18 samples yielded

18 results, as samples 19 and 20 were too small for analysis. The 훿 Osc range of MSL 742 was

18 19.71‰ to 22.33‰ with an amplitude of 2.62‰ (Table 2 & Figures 20 to 22). 훿 Odw values ranged from -10.55‰ to -13.28‰ with an amplitude of 2.73‰ (Table 2). 훿18O values are comparable to those acquired from tusk dentin by Fisher et al. (2003) at the Mammoth Site.

Dietary patterns in MSL 742 훿13C values were stable throughout the sampled time period, ranging from -22.54‰ to -21.39‰, an amplitude of only 1.15‰ (Table 2 & Figures 20 to 22).

The percent of the diet stemming from C4 plants ranged from 29% to 37% (Table 2). The low amplitude of 훿13C of MSL 742 indicates it consumed a relatively consistent diet across the 2 years of the study.

All 20 87Sr/86Sr samples yielded viable results. Values range from 0.71049 to 0.71193 with an amplitude of 0.00144 (Table 2). When mapped across the 87Sr/86Sr isoscape,

Table 2. Micromill Isotopic Values from MSL 742

Figure 20. Stable isotope results for MSL 742 across the sample span

Figure 21. Results of 훿13C vs 훿18O in MSL 742

Figure 22. Results of 훿13C vs 훿18O in MSL 742 with 훿13C shifted backwards one sample

Figure 23. MSL 742 total potential mobility across all 20 samples

Figure 24. Most common areas of MSL 742’s potential mobility across all 20 samples

Figure 25. Most common areas of MSL 742’s potential mobility across all 20 samples, zoomed into the Southern Black Hills

Discussion

Influence of Hot Springs on Local Isotopic Variability

The sinkhole pond at the Mammoth Site that entrapped these mammoths is believed to have been fed by heated springs. Laury (1980) noted that the lack of diatoms and other diverse invertebrate faunas indicated the pond maintained a year-round temperature of at least 35 degrees C with little to no seasonal temperature fluctuations. The presence of Physa sp. at the site combined with a relative lack of molluscan diversity also points to the presence of a heated body of water; hot bodies of water typically host a less diverse range of mollusks than cool

74 bodies of water (Mead et al. 1994). A stable, heated pond would have remained unfrozen for the entirety of the year, granting year-round access to megafaunal vertebrates.

87Sr/86Sr results indicated that MSL 742 remained within the southern Black Hills for the entirety of the sample period (~ 2 years) (Figure 23). A lack of widespread movement across the region suggests that all ingested drinking water and vegetation was locally sourced. Drinking water acquired from precipitation-fed puddles, ponds, or streams, often shows strong seasonal variation in 훿18O values (Różański et al. 2001). Analysis of 훿18O from a study on Great Lakes mastodons displayed this pattern clearly. The mastodons possessed strong seasonal fluctuations in 훿18O values, with amplitudes ranging from 5.5‰ to 8.6‰, indicating water was sourced from precipitation-fed surface water, not more isotopically stable permanent lakes (Metcalfe and

Longstaffe 2014). Mammoths can also display seasonal 훿18O variation, with Missouri mammoths’ fluctuations reaching amplitudes up to 7‰ and Arizona mammoths reaching up to

5‰ (Metcalfe et al. 2011; Widga et al. 2020).

The largest range of seasonal 훿18O values within precipitation occurs between the months of January and July. In the modern Black Hills, this seasonal amplitude can be as high as 24‰

(IAEA 2001). The endpoints of the annual 훿18O range reflect seasonal, continental, and altitudinal effects on 훿18O (Gat et al. 2001). While annually resolved 훿18O samples (i.e., bulk samples) can result in the cancelling out of the two end points, seasonally resolved samples (e.g., micro-samples) clearly reflect these larger seasonal variations.

OSL dating of the Mammoth Site places it firmly in MIS-6, during the peak of the

Illinoian Glaciation. Climate reconstructions of the Black Hills during the LGM derived from fossilized mollusks indicate the region was likely cooler than modern, with fewer temperature

75 extremes (Jass et al. 2002). Due to the similarity of glacial conditions, results from Jass et al.

(2002) are assumed to generally reflect climatic conditions in the region during MIS 6. 훿18O results from the bulk samples of all four mammoths can neither confirm nor deny this. Any seasonal fluctuations that may have occurred are lost to an annual average. However, serial enamel micromill samples are seasonally resolved, potentially at the scale of a monthly average.

If extreme weather events occurred, such as intensive droughts, they would be reflected in the serial analyses.

The relative stability of MSL 742’s 훿18O values, with an amplitude of only 2.73‰, points to the artesian springs in the area as the primary sources of drinking water (Figure 20). As an aquifer-fed system that flows through low permeability Precambrian rocks of the Black Hills, it is likely that the water source of the springs is isotopically stable relative to smaller, precipitation-fed surface water bodies (Miller and Driscoll 1998; Huddart et al. 1999). Williams

Lake and Shingobee Lake in Minnesota, located at a more northern latitude to the Black Hills, exhibited lower seasonal amplitude in 훿18O values than samples taken nearby from precipitation- based surface water (Reddy et al. 2006). With constant influx of isotopically stable water into spring-fed ponds, high 훿18O precipitation during the summer and low 훿18O precipitation during the winter would be diluted within the ponds. If extreme seasonality occurred frequently, enriched and depleted values would still stand out amongst the dampened 훿18O samples. This is not seen in MSL 742, supporting the notion that MIS 6 in the Black Hills supported more equable temperatures than modern. The glacial climate of MIS 6 may have also contributed to less evaporation, further limiting the enrichment of 훿18O values within the pond.

In addition, any vegetation growing in and around the springs would be isotopically composed of the same 훿18O of the water, additionally limiting fluctuation in 훿18O values. Local

76 precipitation likely contributed to the water budget as well, with potential aridity limiting the quantity of precipitation. With the hot springs dynamics not fully understood, the 훿18O values of the site are assumed to derive from the spring. In extant elephants, heavy foraging often occurs around stable water sources (van Aarde et al. 2008). Vegetation growing in wetter habitats is often denser and more taxonomically diverse. Due to the sinkhole pond’s warm, stable temperatures, year-round vegetation would be supported around the margins (Laury 1980).

With no evidence of MSL 742 traveling far away from the site, it can be assumed the springs were able to sustain enough year-round vegetation to support a mammoth population.

Thus, the consumption of primarily locally sourced water and vegetation during more equable climatic conditions explains the lack of any strong 훿18O seasonal fluctuations in MSL 742. As

Fisher et al. (2003) mentions, the unique situation of access to hot springs may limit the application of these results to broader temporal and geographic mammoth studies. However, as unique of a site as it is, it fits into the greater landscape of the Black Hills and Great Plains and opens questions into mammoth population dynamics in the region.

MSL 742 Dietary Reconstruction

Vegetation reconstructions of the Black Hills from pollen at the Mammoth Site, at the time of publication considered to be 26,000 years old, describe the landscape as an arid shrub- steppe with few trees (Mead et al. 1994). With the current date placing the site within MIS 6, interpretations of the site remain valid, as they were based on pollen found within the site, only now reflecting an age older than initially thought (Mahan et al. 2016). Due to the topography of the region, there are very few late Pleistocene fossil localities or lacustrine deposits in the Black

Hills (Jass et al. 2002). Most studies that look at the late Pleistocene stem from the southern

Black Hills, which vary drastically, with respect to both climate and vegetation, from the central and northern Black Hills (Jass et al. 2002).

Additional analyses of fossil mollusks from the Nelson-Wittenberg Site (37,900 yr. BP) noted that within the greater, regional grassland-dominated steppe, the southern Black Hills likely contained a mosaic of mesic micro-habitats including woodlands and wet meadows surrounding the various springs (Jass et al. 2002). The presence of various springs within the region drove the micro-habitat mosaic and as such, it is assumed the southern Black Hills contained a similar mosaic during the active deposition phase of the Mammoth Site due to the presence of active springs. In extant elephants, regions with geographic and temporally dynamic vegetation communities are often preferred over areas that are less dynamic (Murwira and

Skidmore 2005). The homogeneity of the surrounding Great Plains in contrast to the southern

Black Hills may have further incentivized the site for mammoth populations.

The presence of MSL 742 in the region for at least two years suggests the local vegetation was accessible year-round and sufficient to support caloric needs. A closer look into the micromilled 훿13C values of MSL 742 uncovers a slight undulation in the transect. Although the range between high and low points is small at 1.15‰, there are evenly spaced high points

(Figure 21). Sample 2 peaks at -21.68‰, sample 10 peaks at -21.39‰, and sample 18 peaks at -

21.65‰ (Figure 21). Each peak is separated by 8 samples. These peaks all relate to periods of relatively greater C4 plant ingestion. C4 plants are known to peak in productivity and abundance in summer, indicating the peaks are likely representative of the summer season (Ode et al. 1980).

C4 abundance is related to heat availability, with higher minimum summer temperatures resulting in increased C4 success (Teeri and Stowe 1976). During the Illinoian glacial period,

78 summer temperatures were cooler than modern, likely limiting the abundance of C4 plants. At its highest, 37% of MSL 742’s diet consisted of C4 plants (Table 2).

When comparing 훿13C values to 훿18O values, there is little association between the two.

Upon closer inspection, there appears to be a correspondence between the two that is offset by one sample. When this offset is accounted for, and dietary 훿13C values are shifted backwards one sample, there is a clear pattern that emerges (Figure 22). It is hypothesized that this lag in dietary

훿13C values behind 훿18O values reflects an ecological lag time. C4 plants exhibit growth after rainfall events (Uno et al. 2020a). After a summer rain, there exists a lag period for the precipitation to soak into the soil, be absorbed by vegetation, and then consumed by mammals.

In the case of MSL 742, any water ingested during or immediately after precipitation events would be instantly reflected within the isotopic record. However, due to the lag time in that same water reaching the vegetation and then the diet, it is possible it would fail to show concurrently with 훿18O. This may be particularly relevant if the consumption of C4 plants following summer blooms accounts for the shift.

Looking at the corresponding 훿18O peaks, all three peaks account for three of the four largest 훿18O values (Figure 22). Typically, higher 훿18O values correspond to warmer temperatures and lower 훿18O correspond to colder temperatures. This would pair the higher, warmer 훿18O peaks with the higher, C4 훿13C peaks. It is noteworthy that the 훿18O transect displays less of an undulating pattern than the 훿13C transect. This can likely be attributed to two reasons. First, isotopically stable water from the hot springs likely dominates the mammoth’s intake, both in drinking water and through vegetation. This dampens the seasonal variation that may be expected in 훿18O. The two lone peaks in samples 1 and 9, which correlate to peaks in

훿13C and reflect warmer conditions, may come from increased influence of precipitation (Figure

22). In winter, the heated springs remained unfrozen, and likely was the dominant source of drinking water during the cold season. In summer, precipitation-fed puddles and springs would be more common, and mammoths may occasionally utilize these sources out of convenience.

This may lead to random, more positive 훿18O signature of ingested water during summer months, particularly in a wet summer. Second, summer rains can spur the onset of large blooms of C4 grasses, leading to increased consumption by mammoths through availability alone. This temporary bloom enriches not only the 훿13C values of the samples, but the 훿18O values, as any precipitation fed C4 blooms will consist of a more enriched 훿18O record. If a wet summer is paired with increased consumption of C4 grasses, it may co-occur with elevated 훿18O values.

If the evenly spaced isotopic peaks in MSL 742 are annual features, enamel formation in

MSL 742 is slower than the typical range for Columbian mammoths (Figure 21). Enamel extension rates in Columbian mammoths average 13-14 mm per year (Metcalfe and Longstaffe

2012), but in MSL 742, one year would account for 8 mm. In winter, mammoths have been known to limit growth of enamel to focus energy elsewhere (Widga et al. 2020). For MSL 742, although climate conditions were more less seasonal than modern, the overall climate was cooler, as the Black Hills were near the margin of the Illinoian Glaciation. If MSL 742 experienced seasonal nutritional stress, a reduced annual growth of 8mm may be possible.

Geographical Influences

Across the 2-year sampling period, 87Sr/86Sr values indicate MSL 742 maintained strong site fidelity and likely remained within 30 km of the Mammoth Site. Site fidelity is the tendency for an animal to occupy the same area over a long period of time (Switzer 1993). Mapped

87Sr/86Sr values indicate MSL 742 likely lived in and around the southern foothills of the Black 80

Hills, rarely venturing into the more rugged core of the Black Hills or venturing out onto the

Great Plains. The higher 87Sr/86Sr ratios found in the core of the Black Hills were not present in any of the samples. The center of the Black Hills contains very old, exposed rocks possessing very high 87Sr/86Sr ratios. This region of the Black Hills also contains the highest elevations and steepest slopes, with the highest point reaching over 7000 feet.

Extant elephants often avoid steep slopes, despite their ability to traverse them. Wall et al. (2006) noted that African elephants avoided steep terrain despite the presence of ideal vegetation on steeper slopes. They also pointed out the massive energy cost for an elephant to ascend slopes, which can reach as high as 2500% the energy cost of walking on a level surface

(Wall et al. 2006). The rugged terrain in the central Black Hills likely discouraged mammoths from travelling into more rugged areas.

The presence of sloped terrain effects extant elephant populations beyond energy expenditure. In modern elephant populations, male and female elephants utilize the landscape differently in response to abiotic and biotic factors (Shannon et al. 2010). In Kruger National

Park, South Africa, female elephants preferred gentle or level terrain in contrast to males, who clustered in regions with intermediate slopes, avoiding both flat and steep terrain (de Knegt et al.

2011). Female elephants, who live together in maternal herds with juvenile offspring, possess higher relative nutritional demands in contrast to males, further limiting the cost-benefit of scaling slopes (Shannon et al. 2010). In addition, the presence of juveniles in maternal herds makes landscape use of slopes more difficult for females. Bull elephants can tolerate higher variation in food quality and spend more time foraging in the same location (Stokke and Toit

2000). Smit et al. (2007) noted that in almost any studied elephant population, bull areas existed that consisted of only bulls. Mixed herds and maternal herds avoided these areas completely.

The high density of slopes in the southern Black Hills may play a role in the lack of female and juvenile mammoths at the Mammoth Site. When identifying slopes greater than 10°, the entire town of Hot Springs is separated from the Great Plains by a ‘wall’ of more rugged topography (Figure 27). When targeting any slopes of 5° or higher, it quickly becomes apparent the region is inaccessible without traversing these slopes (Figure 26). Slopes of 15° or higher would have been extremely difficult for mammoths to traverse. Although not widespread, these steep slopes cover the southern and eastern flank of the Mammoth Site (Figure 28). The energy cost for mammoths to utilize this landscape is high, as flat regions in the Black Hills are limited.

When MSL 742’s potential mobility is mapped alongside slopes, it is evident that MSL 742 potentially avoided the steepest slopes but was utilizing lesser slopes (Figures 29 and 30). This method assumes the slopes present in the region today are similar to those that were present when the Mammoth Site was active. It is possible, due to the erosional nature of the limestone in the region, that the slopes during MIS 6 were steeper and taller than they were today. If that were true, parts of the region would be even less accessible to mammoths than they are today.

However, given the prominence of bedrock within the region, it is likely the slopes would have been very comparable to modern.

Furthermore, the Mammoth Site may represent a ‘bull area’ for Columbian mammoths, explaining the lack of females and juveniles at the site. The rugged topography of the region likely discouraged matriarchal and mixed herds and led to their avoidance of the region. The site is not a mass-kill site, but an accretional death assemblage that resulted from individual deaths on a scale of decades, centuries, to even millennia. If females were in the region, it is likely fossil remains would have been uncovered at the site.

The Mammoth Site

Figure 26. Slopes within the Black Hills at or above 5°

The Mammoth Site

Figure 27. Slopes within the Black Hills at or above 10°

The Mammoth Site

Figure 28. Slopes within the Black Hills at or above 15°

Figure 29. The total mobility of MSL 742 (displayed from most to least common across all 20 samples) alongside all slopes at or greater than 5° within the Black Hills

Figure 30. The total mobility of MSL 742 (displayed from most to least common across all 20 samples) alongside all slopes at or greater than 10° within the Black Hills

Implications for Mammoth Paleoecology

MSL 742 maintained strong site fidelity to the southern Black Hills for at least 2 years.

The presence of a stable, heated water source likely enabled this behavior, providing year-round forage and access to drinking water. The lack of seasonal migrations in and out of the site highlights the importance of the heated springs in the region.

This study suggests a new hypothesis for the dominance of male sub-adult and adult

Columbian mammoths at the site. The current understanding has suggested that the lack of females and juveniles can be attributed to the societal knowledge contained in maternal herds.

Extant elephants are highly intelligent and carry in-depth local knowledge about the landscape

(Archie et al. 2006). In this model, young males, who leave the herd as they mature (Lee 1986), are drawn to the sinkhole pond without the herd-based knowledge of its danger. The bold and naïve males thus enter the pond, realizing too late they are unable to escape, leading to their eventual drowning. Pečnerová et al. (2017) sexed 98 mammoth remains and determined a significant skew towards males, supporting the notion that sub-adult males were more prone to natural entrapment without the knowledge provided by herd social structure.

This study suggests the region supported male-dominated populations across the time of deposition at the site. The rugged terrain in the region likely discouraged matriarchal herds from foraging around the Hot Springs. The less rugged Great Plains that surround the Black Hills would be much more energetically efficient for matriarchal herds, especially those with juveniles. The nutritional demands of lactating and breeding female elephants is high, and the energy lost to constantly traversing slopes would present a large problem. Bull elephants do not share the same nutritional constraints and can forage among lesser quality vegetation than females. In addition to potentially carrying generational knowledge of the dangers of the hot springs, it is suggested that local topography limited the usage of the southern Black Hills by matriarchal herds.

The ability for mammoths to remain in the southern Black Hills year-round indicates the region likely contained a diverse mosaic of vegetational communities. As a keystone species, the presence of mammoths year-round likely opened niches for a diverse array of mammalian species in the region, some of which have also been uncovered at the site. The unique ecology of the Mammoth Site, created by the presence of hot springs, allows for an in-depth examination of how mammoths utilized diverse landscapes in comparison to other regions of the continent.

Conclusion

The Mammoth Site remains a major focus for North American mammoth research, as one of the densest and most complete mammoth fossil localities in the North America.

Understanding the ecological context of this fossil assemblage is key to better understanding how mammoths utilized the landscape, as well as the impact of regional topography, water body distribution, and vegetation communities influenced mammoth behavior. Isotopic sampling of four Mammoth Site molars provided a deeper look into the paleoecology of Columbian mammoths at the site.

Analysis of mammoth mobility via 87Sr/86Sr confirmed that MSL 742 possessed strong site fidelity, living in the southern Black Hills for at least 2 years. Seasonal changes in climate did not force MSL 742 to leave the region or alter mobility patterns in any discernible way.

Dietary analyses indicate ingestion of forage with a very limited range of 훿13C values, highlighting the consistency and stability of the diet over the 2-year study stretch. The diet primarily consisted of C3 plants, with subtle increases in C4 forage during summer months.

Dietary signals appeared to lag ~4 weeks behind 훿18O signals in MSL 742, producing evidence of a possible ecological lag on the landscape. 훿18O results confirmed the climate was likely cooler, but less seasonal, than modern conditions. The relatively low-amplitude 훿18O values likely resulted from heavy intake of drinking water and vegetation sourced from an isotopically invariable spring-fed pond. Sporadic intake of precipitation-fed water created low amplitude spikes in 훿18O values during summer months.

The lack of female or juvenile mammoths at the Mammoth Site is suggested to be the result of an unfavorable, sloped landscape in and around the southern Black Hills. With the deposition period of the site occurring during MIS 6, local slopes in the region were potentially 89 steeper than they were today. Limestone has high erosion rates, and modern slopes may reflect an eroded version of slopes present when the site was active. Such sloped areas are avoided by extant maternal herds and utilized by modern male elephants. In addition, it is very common in regions containing elephant populations to have areas described as ‘bull areas’, where the only elephants in the region are males. This provides a new hypothesis for the male-dominated mammoths at the Mammoth Site.

REFERENCES van Aarde R, Ferreira S, Jackson T, Page B, Ernst Y, Gough K, Guldemond R, Junker J, Olivier

P, Ott T, et al. 2008. Elephant population biology and ecology. In: Elephant

Management: A Scientific Assessment for South Africa. p. 84–145.

Agenbroad LD. 1994a. A Brief History of the Hot Springs Mammoth Site. In: Agenbroad LD,

Mead JI, editors. The Hot Springs Mammoth Site: a decade of field and laboratory

research in paleontology, geology, and paleoecology. Mammoth Site of Hot Springs,

South Dakota, Inc. p. 4–14.

Agenbroad LD. 1994b. Geology, Hydrology, and Excavation of the Site. In: Agenbroad LD,

Mead JI, editors. The Hot Springs Mammoth Site: a decade of field and laboratory

research in paleontology, geology, and paleoecology. Mammoth Site of Hot Springs,

South Dakota, Inc. p. 15–27.

Agenbroad LD. 1994c. Taxonomy of North American Mammuthus and Biometrics of the Hot

Springs Mammoths. In: Agenbroad LD, Mead JI, editors. The Hot Springs Mammoth

Site: a decade of field and laboratory research in paleontology, geology, and

paleoecology. Mammoth Site of Hot Springs, South Dakota, Inc. p. 158–207.

Agenbroad LD. 2003. New absolute dates and comparisons for California’s Mammuthus exilis.

Deinsea. 9:1–16.

Agenbroad LD. 2005. North American Proboscideans: Mammoths: The state of Knowledge,

2003. Quaternary International. 126–128:73–92. doi:10.1016/j.quaint.2004.04.016.

Anders U, Koenigswald W. 2013. Increasing and decreasing functional area of the dentition

(FAD) of Mammuthus primigenius. Paläontologische Zeitschrift. 87.

doi:10.1007/s12542-013-0168-2.

Archie E, Moss C, Alberts S. 2006. The ties that bind: Genetic relatedness predicts the fission

and fusion of social groups in wild African elephants. Proceedings Biological sciences /

The Royal Society. 273:513–22. doi:10.1098/rspb.2005.3361.

Ayliffe LK, Lister AM, Chivas AR. 1992. The preservation of glacial-interglacial climatic

signatures in the oxygen isotopes of elephant skeletal phosphate. Palaeogeography,

Palaeoclimatology, Palaeoecology. 99(3):179–191. doi:10.1016/0031-0182(92)90014-V.

Barnes RFW. 1983. Effects of Elephant Browsing on Woodlands in a Tanzanian National Park:

Measurements, Models and Management. The Journal of Applied Ecology. 20(2):521.

doi:10.2307/2403524.

Barnosky AD, Koch PL, Feranec RS, Wing SL, Shabel AB. 2004. Assessing the Causes of Late

Pleistocene Extinctions on the Continents. Science. 306(5693):70.

doi:10.1126/science.1101476.

Bataille CP, Bowen GJ. 2012. Mapping 87Sr/86Sr variations in bedrock and water for large scale

provenance studies. Chemical Geology. 304–305:39–52.

doi:10.1016/j.chemgeo.2012.01.028.

Bataille CP, Brennan SR, Hartmann J, Moosdorf N, Wooller MJ, Bowen GJ. 2014. A

geostatistical framework for predicting variations in strontium concentrations and isotope

ratios in Alaskan rivers. Chemical Geology. 389:1–15.

doi:10.1016/j.chemgeo.2014.08.030.

Bataille CP, von Holstein ICC, Laffoon JE, Willmes M, Liu X-M, Davies GR. 2018. A

bioavailable strontium isoscape for Western Europe: A machine learning approach.

Cotton J, editor. PLoS ONE. 13(5):e0197386. doi:10.1371/journal.pone.0197386.

Batchelor CL, Margold M, Krapp M, Murton DK, Dalton AS, Gibbard PL, Stokes CR, Murton

JB, Manica A. 2019. The configuration of Northern Hemisphere ice sheets through the

Quaternary. Nat Commun. 10(1):3713–3713. doi:10.1038/s41467-019-11601-2.

Beard B, Johnson C. 2000. Strontium Isotope Composition of Skeletal Material Can Determine

the Birth Place and Geographic Mobility of Humans and Animals. Journal of forensic

sciences. 45:1049–61. doi:10.1520/JFS14829J.

Bender MM. 1971. Variations in the 13C/12C ratios of plants in relation to the pathway of

photosynthetic carbon dioxide fixation. Phytochemistry. 10(6):1239–1244.

doi:10.1016/S0031-9422(00)84324-1.

Bentley AR. 2006. Strontium Isotopes from the Earth to the Archaeological Skeleton: A Review.

J Archaeol Method Theory. 13(3):135–187. doi:10.1007/s10816-006-9009-x.

Breiman L. 2001. Random Forests. Machine Learning. 45(1):5–32.

doi:10.1023/A:1010933404324.

Broughton J, Weitzel E. 2018. Population reconstructions for humans and megafauna suggest

mixed causes for North American Pleistocene extinctions. Nature Communications. 9.

doi:10.1038/s41467-018-07897-1.

Bryant DJ, Froelich PN. 1995. A model of oxygen isotope fractionation in body water of large

mammals. Geochimica et Cosmochimica Acta. 59(21):4523–4537. doi:10.1016/0016-

7037(95)00250-4.

Bryant DJ, Koch PL, Froelich PN, Showers WJ, Genna BJ. 1996. Oxygen isotope partitioning

between phosphate and carbonate in mammalian apatite. Geochimica et Cosmochimica

Acta. 60(24):5145–5148. doi:10.1016/S0016-7037(96)00308-0.

Burke WH, Denison RE, Hetherington EA, Koepnick RB, Nelson HF, Otto JB. 1982. Variation

of seawater 87Sr/86Sr throughout Phanerozoic time. Geology. 10(10):516–519.

doi:10.1130/0091-7613(1982)10<516:VOSSTP>2.0.CO;2.

Capo RC, Stewart BW, Chadwick OA. 1998. Strontium isotopes as tracers of ecosystem

processes: theory and methods. Geoderma. 82(1):197–225. doi:10.1016/S0016-

7061(97)00102-X.

Cerling TE, Ehleringer JR. 2000. Welcome to the C4 World. The Paleontological Society Papers.

6:273–286. doi:10.1017/S1089332600000802.

Cerling TE, Harris JM. 1999. Carbon isotope fractionation between diet and bioapatite in

ungulate mammals and implications for ecological and paleoecological studies.

Oecologia. 120(3):347–363. doi:10.1007/s004420050868.

Claesson S, Baleka S, Hofreiter M, Widga C. 2017. The contribution of Late Pleistocene

megafauna finds to submerged archaeology and the interpretation of ancient coastal

landscapes. Journal of Archaeological Science: Reports. 15:290–298.

doi:10.1016/j.jasrep.2017.08.007.

Connin SL, Betancourt J, Quade J. 1998. Late Pleistocene C 4 Plant Dominance and Summer

Rainfall in the Southwestern United States from Isotopic Study of Herbivore Teeth. Quat

res. 50(2):179–193. doi:10.1006/qres.1998.1986.

Coplen TB. 1995. Reporting of stable hydrogen, carbon, and oxygen isotopic abundances.

Geothermics. 24(5):707–712. doi:10.1016/0375-6505(95)00024-0.

Coplen TB, Brand W, Gehre M, Gröning M, Meijer HAJ, Toman B, Verkouteren R. 2006. New

guidelines for 훿13C measurements. Anal Chem. 78.

Czaplewski N, Mead J. 1994. Late Pleistocene small mammals from Hot Springs Mammoth Site,

South Dakota. The Hot Springs Mammoth Site: a decade of field and laboratory research

in paleontology, geology, and paleoecology The Mammoth Site of Hot Springs, SD.:136–

149.

Daux V, Lécuyer C, Héran M-A, Amiot R, Simon L, Fourel F, Martineau F, Lynnerup N,

Reychler H, Escarguel G. 2008. Oxygen isotope fractionation between human phosphate

and water revisited. Journal of Human Evolution. 55(6):1138–1147.

doi:10.1016/j.jhevol.2008.06.006.

Davis M, Pineda‐Munoz S. 2016. The temporal scale of diet and dietary proxies. Ecol Evol.

6(6):1883–1897. doi:10.1002/ece3.2054.

Dawson TE, Mambelli S, Plamboeck AH, Templer PH, Tu KP. 2002. Stable Isotopes in Plant

Ecology. Annu Rev Ecol Syst. 33(1):507–559.

doi:10.1146/annurev.ecolsys.33.020602.095451.

De Boer WF, Ntumi CP, Correia AU, Mafuca JM. 2000. Diet and distribution of elephant in the

Maputo Elephant Reserve, Mozambique: Diet and distribution of elephants. African

Journal of Ecology. 38(3):188–201. doi:10.1046/j.1365-2028.2000.00243.x.

Dirks W, Bromage TG, Agenbroad LD. 2012. The duration and rate of molar plate formation in

Palaeoloxodon cypriotes and Mammuthus columbi from dental histology. Quaternary

International. 255:79–85. doi:10.1016/j.quaint.2011.11.002.

Dworkin SI, Nordt L, Atchley S. 2005. Determining terrestrial paleotemperatures using the

oxygen isotopic composition of pedogenic carbonate. Earth and Planetary Science

Letters. 237(1):56–68. doi:10.1016/j.epsl.2005.06.054.

Emery-Wetherell MM, McHorse BK, Byrd Davis E. 2017. Spatially explicit analysis sheds new

light on the Pleistocene megafaunal extinction in North America. Paleobiology.

43(4):642–655. doi:10.1017/pab.2017.15.

Enk J, Devault A, Widga C, Saunders J, Szpak P, Southon J, Rouillard J-M, Shapiro B, Golding

GB, Zazula G, et al. 2016. Mammuthus Population Dynamics in Late Pleistocene North

America: Divergence, Phylogeography, and Introgression. Front Ecol Evol. 4.

doi:10.3389/fevo.2016.00042.

Esker D, Forman SL, Widga C, Walker J, Andrew J. 2019. Home range of the Columbian

mammoths (Mammuthus columbi) and grazing herbivores from the Waco Mammoth

National Monument, (Texas, USA) based on strontium isotope ratios from tooth enamel

bioapatite. Palaeogeography, Palaeoclimatology, Palaeoecology. 534:109291.

doi:10.1016/j.palaeo.2019.109291.

Fernando P, Wikramanayake ED, Janaka HK, Jayasinghe LKA, Gunawardena M, Kotagama

SW, Weerakoon D, Pastorini J. 2008. Ranging behavior of the Asian elephant in Sri

Lanka. Mammalian Biology. 73(1):2–13. doi:10.1016/j.mambio.2007.07.007.

Ferretti M. 2003. Structure and evolution of mammoth molar enamel. Acta Palaeontologica

Polonica. 48.

Fick SE, Hijmans RJ. 2017. WorldClim 2: new 1‐km spatial resolution climate surfaces for

global land areas. Int J Climatol. 37(12):4302–4315. doi:10.1002/joc.5086.

Field CR. 1971. Elephant ecology in the Queen Elizabeth National Park, Uganda. African

Journal of Ecology. 9(1):99–123. doi:10.1111/j.1365-2028.1971.tb00223.x.

Fisher DC, Fox DL, Agenbroad LD. 2003. Tusk growth rate and season of death of Mammuthus

columbi from Hot Springs, South Dakota, USA. Deinsea. 9:117–134.

Forman SL, Arthur Bettis E, Kemmis TJ, Miller BB. 1992. Chronologic evidence for multiple

periods of loess deposition during the Late Pleistocene in the Missouri and Mississippi

River Valley, United States: Implications for the activity of the Laurentide ice sheet.

Palaeogeography, Palaeoclimatology, Palaeoecology. 93(1–2):71–83. doi:10.1016/0031-

0182(92)90184-7.

Fox DL, Fisher DC, Vartanyan S, Tikhonov AN, Mol D, Buigues B. 2007. Paleoclimatic

implications of oxygen isotopic variation in late Pleistocene and Holocene tusks of

Mammuthus primigenius from northern Eurasia. Quaternary International. 169–170:154–

165. doi:10.1016/j.quaint.2006.09.001.

Gat JR, Mook WG, Meijer HAJ. 2001. Atmospheric Water. International Atomic Energy

Agency and United Nations Educational, Scientific and Cultural Organization

(Environmental Isotopes in the Hydrological Cycle: Principles and Applications).

Gheerbrant E. 2009. Paleocene emergence of elephant relatives and the rapid radiation of

African ungulates. Proceedings of the National Academy of Sciences. 106(26):10717–

10721. doi:10.1073/pnas.0900251106.

Graustein WC. 1989. 87Sr/86Sr Ratios Measure the Sources and Flow of Strontium in Terrestrial

Ecosystems. In: Rundel PW, Ehleringer JR, Nagy KA, editors. Stable Isotopes in

Ecological Research. Vol. 68. New York, NY: Springer New York. (Billings WD, Golley

F, Lange OL, Olson JS, Remmert H, editors. ). p. 491–512. [accessed 2020 Jan 1].

http://link.springer.com/10.1007/978-1-4612-3498-2_28.

Grayson D. 2007. Deciphering North American Pleistocene Extinctions. Journal of

Anthropological Research. 63. doi:10.3998/jar.0521004.0063.205.

Grayson DK, Meltzer DJ. 2015. Revisiting Paleoindian exploitation of extinct North American

mammals. Journal of Archaeological Science. 56:177–193.

doi:10.1016/j.jas.2015.02.009.

Grootes PM, Stuiver M. 1997. Oxygen 18/16 variability in Greenland snow and ice with 10−3- to

105-year time resolution. Journal of Geophysical Research: Oceans. 102(C12):26455–

26470. doi:10.1029/97JC00880.

Grootes PM, Stuiver M, White JWC, Johnsen S, Jouzel J. 1993. Comparison of oxygen isotope

records from the GISP2 and GRIP Greenland ice cores. Nature. 366(6455):552–554.

doi:10.1038/366552a0.

Haynes G. 1991. Mammoths, mastodonts, and elephants: biology, behavior, and the fossil

record. Cambridge : New York: Cambridge University Press.

Haynes G. 2002. The Catastrophic Extinction of North American Mammoths and Mastodonts.

World Archaeology. 33:391–416. doi:10.1080/00438240120107440.

Hengl T, Mendes de Jesus J, Heuvelink GBM, Ruiperez Gonzalez M, Kilibarda M, Blagotić A,

Shangguan W, Wright MN, Geng X, Bauer-Marschallinger B, et al. 2017.

SoilGrids250m: Global gridded soil information based on machine learning. Bond-

Lamberty B, editor. PLoS ONE. 12(2):e0169748. doi:10.1371/journal.pone.0169748.

Higgins P. 2018. Isotope Ecology from Biominerals. In: Croft DA, Su DF, Simpson SW, editors.

Methods in Paleoecology: Reconstructing Cenozoic Terrestrial Environments and

Ecological Communities. Cham: Springer International Publishing. p. 99–120.

https://doi.org/10.1007/978-3-319-94265-0_7.

Holmgren C, Norris J, Betancourt J. 2007. Inferences about winter temperatures and summer

rains from the Late Quaternary record of C4 perennial grasses and C3 desert shrubs in the

northern Chihuahuan Desert. Journal of Quaternary Science. 22:141–161.

doi:10.1002/jqs.1023.

Hoppe K, Koch P. 2006. The Biogeochemistry of the Aucilla River Fauna. In: Topics in

Geobiology. Vol. 26. p. 379–401.

Hoppe KA, Koch PL. 2007. Reconstructing the migration patterns of late Pleistocene mammals

from northern Florida, USA. Quaternary Research. 68(3):347–352.

doi:10.1016/j.yqres.2007.08.001.

Horton JD. 2017. The State Geologic Map Compilation (SGMC) Geodatabase of the

Conterminous United States. doi:10.5066/F7WH2N65.

Huddart PA, Longstaffe FJ, Crowe AS. 1999. δD and δ18O evidence for inputs to groundwater at

a wetland coastal boundary in the southern Great Lakes region of Canada. Journal of

Hydrology. 214(1–4):18–31. doi:10.1016/S0022-1694(98)00226-1.

Iacumin P, Bocherens H, Mariotti A, Longinelli A. 1996. Oxygen isotope analyses of co-existing

carbonate and phosphate in biogenic apatite: a way to monitor diagenetic alteration of

bone phosphate? Earth and Planetary Science Letters. 142(1–2):1–6. doi:10.1016/0012-

821X(96)00093-3.

Jass C, Mead J, Morrison A, Agenbroad L. 2002. Late Pleistocene mollusks from the southern

Black Hills, South Dakota. Western North American Naturalist. 62:129–140.

Kendall C, Coplen T. 2001. Distribution of oxygen-18 and deuterium in river waters across the

United States. Hydrological Processes. 15:1363–1393. doi:10.1002/hyp.217. de Knegt HJ, van Langevelde F, Skidmore AK, Delsink A, Slotow R, Henley S, Bucini G, de

Boer WF, Coughenour MB, Grant CC, et al. 2011. The spatial scaling of habitat selection

by African elephants: Scaling habitat selection by elephants. Journal of Animal Ecology.

80(1):270–281. doi:10.1111/j.1365-2656.2010.01764.x.

Koch PL, Barnosky AD. 2006. Late Quaternary Extinctions: State of the Debate. Annual Review

of Ecology, Evolution, and Systematics. 37:215–250.

100

Koch PL, Hoppe KA, Webb SD. 1998. The isotopic ecology of late Pleistocene mammals in

North America: Part 1. Florida. Chemical Geology. 152(1):119–138. doi:10.1016/S0009-

2541(98)00101-6.

Koch PL, Tuross N, Fogel ML. 1997. The Effects of Sample Treatment and Diagenesis on the

Isotopic Integrity of Carbonate in Biogenic Hydroxylapatite. Journal of Archaeological

Science. 24(5):417–429. doi:10.1006/jasc.1996.0126.

Kozawa Y, Mishima H, Suzuki K, Ferguson M. 2002. Dental Formula of Elephant by the

Development of Tooth Germ.

Laury RL. 1980. Paleoenvironment of a late Quaternary mammoth-bearing sinkhole deposit, Hot

Springs, South Dakota. GSA Bulletin. 91(8):465–475. doi:10.1130/0016-

7606(1980)91<465:POALQM>2.0.CO;2.

Laws RM. 1966. Age Criteria for the African Elephant. African Journal of Ecology. 4(1):1–37.

doi:10.1111/j.1365-2028.1966.tb00878.x.

Lee P. 1986. Early social development in African elephants. National Geographic Research. 2.

Lister A. 2014. Mammoths and Mastodons of the Ice Age.

Lister A, Essen H. 2003 Jan 1. Mammuthus rumanus (Stefanescu), the earliest mammoth in

Europe. Advances in Palaeontology ‘Hen to Panta’ Romanian Academy.:47–52.

Lister AM, Agenbroad LD. 1994. Gender Determination of the Hot Springs Mammoths. In:

Agenbroad LD, Mead JI, editors. The Hot Springs Mammoth Site: a decade of field and

laboratory research in paleontology, geology, and paleoecology. Mammoth Site of Hot

Springs, South Dakota, Inc. p. 208–214.

101

Lister AM, Sher AV. 2015. Evolution and dispersal of mammoths across the Northern

Hemisphere. Science. 350(6262):805. doi:10.1126/science.aac5660.

Longinelli A. 1984. Oxygen isotopes in mammal bone phosphate: A new tool for

paleohydrological and paleoclimatological research? Geochimica et Cosmochimica Acta.

48(2):385–390. doi:10.1016/0016-7037(84)90259-X.

Maglio VJ. 1973. Origin and Evolution of the Elephantidae. Transactions of the American

Philosophical Society. 63(3):1–149. doi:10.2307/1006229.

Mahan S, Hanson P, Mead J, Holen S, Wilkins W. 2016. The Mammoth Site at Hot Springs,

South Dakota: Use of OSL Dating to Calibrate the Sinkhole Time Machine.

Marshall J, Brooks JR, Lajtha K. 2008. Sources of Variation in the Stable Isotopic Composition

of Plants. In: Stable Isotopes in Ecology and Environmental Science. p. 22–60.

Martin PS. 1973. The Discovery of America. Science. 179(4077):969.

doi:10.1126/science.179.4077.969.

McCullagh K. 1969. The Growth and Nutrition of the African Elephant II. The chemical nature

of the diet. African Journal of Ecology. 7(1):91–97. doi:10.1111/j.1365-

2028.1969.tb01197.x.

Mead JI, Hevly RH, Agenbroad LD. 1994. Late Pleistocene Invertebrate and Plant Remains,

Mammoth Site, Black Hills, South Dakota. In: Agenbroad LD, Mead JI, editors. The Hot

Springs Mammoth Site: a decade of field and laboratory research in paleontology,

geology, and paleoecology. Mammoth Site of Hot Springs, South Dakota, Inc. p. 117–

135.

102

Metcalfe J, Longstaffe F, Ballenger J, Haynes C. 2011. Isotopic paleoecology of Clovis

mammoths from Arizona. Proceedings of the National Academy of Sciences of the

United States of America. 108:17916–20. doi:10.1073/pnas.1113881108.

Metcalfe JZ, Longstaffe FJ. 2012. Mammoth tooth enamel growth rates inferred from stable

isotope analysis and histology. Quaternary Research. 77(3):424–432.

doi:10.1016/j.yqres.2012.02.002.

Metcalfe JZ, Longstaffe FJ. 2014. Environmental change and seasonal behavior of mastodons in

the Great Lakes region inferred from stable isotope analysis. Quaternary Research.

82(2):366–377. doi:10.1016/j.yqres.2014.07.002.

Metcalfe JZ, Longstaffe FJ, Hodgins G. 2013. Proboscideans and paleoenvironments of the

Pleistocene Great Lakes: landscape, vegetation, and stable isotopes. Quaternary Science

Reviews. 76:102–113. doi:10.1016/j.quascirev.2013.07.004.

Miller LD, Driscoll DG. 1998. Streamflow characteristics for the Black Hills of South Dakota,

through water year 1993. Water-Resources Investigations Report Report No.: 97–4288.

http://pubs.er.usgs.gov/publication/wri974288.

Murwira A, Skidmore AK. 2005. The response of elephants to the spatial heterogeneity of

vegetation in a Southern African agricultural landscape. Landscape Ecol. 20(2):217–234.

doi:10.1007/s10980-004-3159-6.

Ode DJ, Tieszen LL, Lerman JC. 1980. The Seasonal Contribution of C3 and C4 Plant Species to

Primary Production in a Mixed Prairie. Ecology. 61(6):1304–1311. doi:10.2307/1939038.

103

Osborn F. 2004. The concept of home range in relation to elephants in Africa. Pachyderm.

37:37–44.

Paley RGT, Kerley GIH. 1998. The winter diet of elephant in Eastern Cape Subtropical Thicket,

Addo Elephant National Park. Koedoe. 41(1):37–45. doi:10.4102/koedoe.v41i1.244.

Pasenko MR, Schubert BW. 2004. Mammuthus jeffersonii (Proboscidea, Mammalia) from

Northern Illinois. PaleoBios. 24(3):19–24.

Pečnerová P, Díez-del-Molino D, Dussex N, Feuerborn T, von Seth J, van der Plicht J, Nikolskiy

P, Tikhonov A, Vartanyan S, Dalén L. 2017. Genome-Based Sexing Provides Clues

about Behavior and Social Structure in the Woolly Mammoth. Current Biology.

27(22):3505-3510.e3. doi:10.1016/j.cub.2017.09.064.

18 18 Pellegrini M, Lee-Thorp JA, Donahue RE. 2011. Exploring the variation of the δ Op and δ Oc

relationship in enamel increments. Palaeogeography, Palaeoclimatology, Palaeoecology.

310(1):71–83. doi:10.1016/j.palaeo.2011.02.023.

Pelletier JD, Broxton PD, Hazenberg P, Zeng X, Troch PA, Niu G, Williams Z, Brunke MA,

Gochis D. 2016. A gridded global data set of soil, intact regolith, and sedimentary deposit

thicknesses for regional and global land surface modeling. J Adv Model Earth Syst.

8(1):41–65. doi:10.1002/2015MS000526.

Pineda-Munoz S, Jukar AM, Tóth AB, Fraser D, Du A, Barr WA, Amatangelo KL, Balk MA,

Behrensmeyer AK, Blois J, et al. 2021. Body mass-related changes in mammal

community assembly patterns during the late Quaternary of North America. Ecography.

44(1):56–66. doi:10.1111/ecog.05027.

104

Post DM. 2002. Using Stable Isotopes to Estimate Trophic Position: Models, Methods, and

Assumptions. Ecology. 83(3):703–718. doi:10.1890/0012-

9658(2002)083[0703:USITET]2.0.CO;2.

Quade J, Cerlinga TE, Barry JC, Morgan ME, Pilbeam DR, Chivas AR, Lee-Thorp JA, van der

Merwe NJ. 1992. A 16-Ma record of paleodiet using carbon and oxygen isotopes in fossil

teeth from Pakistan. Chemical Geology: Isotope Geoscience section. 94(3):183–192.

doi:10.1016/0168-9622(92)90011-X.

Reddy MM, Schuster P, Kendall C, Reddy MB. 2006. Characterization of surface and ground

water δ18O seasonal variation and its use for estimating groundwater residence times.

Hydrol Process. 20(8):1753–1772. doi:10.1002/hyp.5953.

Rivals F, Mihlbachler MC, Solounias N, Mol D, Semprebon GM, de Vos J, Kalthoff DC. 2010.

Palaeoecology of the Mammoth Steppe fauna from the late Pleistocene of the North Sea

and Alaska: Separating species preferences from geographic influence in paleoecological

dental wear analysis. Palaeogeography, Palaeoclimatology, Palaeoecology. 286(1):42–54.

doi:10.1016/j.palaeo.2009.12.002.

Różański K, Froehlich K, Mook WG. 2001. Surface Water. International Atomic Energy Agency

and United Nations Educational, Scientific and Cultural Organization (Environmental

Isotopes in the Hydrological Cycle: Principles and Applications).

Saunders JJ, Grimm EC, Widga CC, Campbell GD, Curry BB, Grimley DA, Hanson PR,

McCullum JP, Oliver JS, Treworgy JD. 2010. Paradigms and proboscideans in the

southern Great Lakes region, USA. Quaternary International. 217(1):175–187.

doi:10.1016/j.quaint.2009.07.031.

105

Semprebon GM, Rivals F, Fahlke JM, Sanders WJ, Lister AM, Göhlich UB. 2016. Dietary

reconstruction of pygmy mammoths from Santa Rosa Island of California. Quaternary

International. 406:123–136. doi:10.1016/j.quaint.2015.10.120.

Shannon G, Page B, Duffy K, Slotow R. 2010. The ranging behaviour of a large sexually

dimorphic herbivore in response to seasonal and annual environmental variation. Austral

Ecology. 35:731–742. doi:10.1111/j.1442-9993.2009.02080.x.

Shoshani J. 1998. Understanding proboscidean evolution: a formidable task. Trends in Ecology

& Evolution. 13(12):480–487. doi:10.1016/S0169-5347(98)01491-8.

Sima A, Paul A, Schulz M, Oerlemans J. 2006. Modeling the oxygen-isotopic composition of the

North American Ice Sheet and its effect on the isotopic composition of the ocean during

the last glacial cycle. Geophysical Research Letters. 33(15). doi:10.1029/2006GL026923.

Sponheimer M, Lee-Thorp JA. 1999. Oxygen Isotopes in Enamel Carbonate and their Ecological

Significance. Journal of Archaeological Science. 26(6):723–728.

doi:10.1006/jasc.1998.0388.

Sponheimer M, Robinson T, Ayliffe L, Passey B, Roeder B, Shipley L, Lopez E, Cerling T,

Dearing D, Ehleringer J. 2003. An experimental study of carbon-isotope fractionation

between diet, hair, and feces of mammalian herbivores. Can J Zool. 81(5):871–876.

doi:10.1139/z03-066.

Stiff BJ, Hansel AK. 2004. Quaternary Glaciations in Illinois. In: Developments in Quaternary

Sciences. Vol. 2. Elsevier. p. 71–82.

106

Stokke S, Toit JT. 2000. Sex and size related differences in the dry season feeding patterns of

elephants in Chobe National Park, Botswana. Ecography. 23(1):70–80.

doi:10.1111/j.1600-0587.2000.tb00262.x.

Suga S. 1989. Enamel Hypomineralization Viewed From the Pattern of Progressive

Mineralization of Human and Monkey Developing Enamel. Adv Dent Res. 3(2):188–

198. doi:10.1177/08959374890030021901.

Switzer PV. 1993. Site fidelity in predictable and unpredictable habitats. Evol Ecol. 7(6):533–

555. doi:10.1007/BF01237820.

Teeri JA, Stowe LG. 1976. Climatic patterns and the distribution of C4 grasses in North

America. Oecologia. 23(1):1–12. doi:10.1007/BF00351210.

Thorp JL, Van der Merwe NJ. 1987. Carbon isotope analysis of fossil bone apatite. South

African Journal of Science. 83(11):712–715.

Uno K, Fisher D, Schuster G, Wittemyer G, Douglas-Hamilton I, Oduor P, Litoroh M, Cerling T.

2020a. High-resolution stable isotope profiles of modern elephant (Loxodonta africana)

tusk dentin and tail hair from Kenya: Implications for identifying seasonal variability in

climate, ecology, and diet in ancient proboscideans. Palaeogeography,

Palaeoclimatology, Palaeoecology. 559:109962. doi:10.1016/j.palaeo.2020.109962.

Uno KT, Fisher DC, Wittemyer G, Douglas-Hamilton I, Carpenter N, Omondi P, Cerling TE.

2020b. Forward and inverse methods for extracting climate and diet information from

stable isotope profiles in proboscidean molars. Quaternary International. 557:92–109.

doi:10.1016/j.quaint.2020.06.030.

107 van der Valk T, Pečnerová P, Díez-del-Molino D, Bergström A, Oppenheimer J, Hartmann S,

Xenikoudakis G, Thomas JA, Dehasque M, Sağlıcan E, et al. 2021 Feb 17. Million-year-

old DNA sheds light on the genomic history of mammoths. Nature. doi:10.1038/s41586-

021-03224-9. [accessed 2021 Feb 26]. http://www.nature.com/articles/s41586-021-

03224-9.

Wall J, Douglas-Hamilton I, Vollrath F. 2006. Elephants avoid costly mountaineering. Current

Biology. 16(14):R527–R529. doi:10.1016/j.cub.2006.06.049.

Wang Y, Cerling TE. 1994. A model of fossil tooth and bone diagenesis: implications for

paleodiet reconstruction from stable isotopes. Palaeogeography, Palaeoclimatology,

Palaeoecology. 107(3):281–289. doi:10.1016/0031-0182(94)90100-7.

Wato YA, Heitkönig IMA, van Wieren SE, Wahungu G, Prins HHT, van Langevelde F. 2016.

Prolonged drought results in starvation of African elephant (Loxodonta africana).

Biological Conservation. 203:89–96. doi:10.1016/j.biocon.2016.09.007.

Wei G, Hu S, Yu K, Hou Y, Li X, Jin C, Wang Y, Zhao J, Wang W. 2010. New materials of the

steppe mammoth, Mammuthus trogontherii, with discussion on the origin and

evolutionary patterns of mammoths. Science China Earth Sciences. 53(7):956–963.

doi:10.1007/s11430-010-4001-4.

Weißbach S, Wegner A, Opel T, Oerter H, Vinther BM, Kipfstuhl S. 2016. Spatial and temporal

oxygen isotope variability in northern Greenland – implications for a new climate record

over the past millennium. Clim Past. 12(2):171–188. doi:10.5194/cp-12-171-2016.

West J, Bowen GJ, Dawson TE, Tu K. 2010. Isoscapes: Understanding Movement, Pattern, and

Process on Earth Through Isotope Mapping.

108

Widga C, Hodgins G, Kolis K, Lengyel S, Saunders J, Walker J, Wanamaker A. 2020. Multi-

scalar Approaches to the stable isotope ecology of late Quaternary Proboscideans in the

Midcontinent.

Widga C, Lengyel S, Saunders J, Hodgins G, Walker J, Wanamaker A. 2017a Feb 1. Late

Pleistocene proboscidean population dynamics in the North American Midcontinent.

Boreas. doi:10.1111/bor.12235.

Widga C, Saunders J, Enk J. 2017b. Reconciling phylogenetic and morphological trends in North

American Mammuthus. Quaternary International. 443:32–39.

doi:10.1016/j.quaint.2017.01.034.

Widga C, Walker JD, Boehm A. 2017c. Variability in Bioavailable 87Sr/86Sr in the North

American Midcontinent. OPEN QUAT. 3(1):4. doi:10.5334/oq.32.

Wittemyer G, Douglas-Hamilton I, Getz W. 2005. The socioecology of elephants: Analysis of

the processes creating multitiered social structures. Animal Behaviour. 69:1357–1371.

doi:10.1016/j.anbehav.2004.08.018.

Zazzo A, Balasse M, Patterson WP. 2006. The reconstruction of mammal individual history:

refining high-resolution isotope record in bovine tooth dentine. Journal of Archaeological

Science. 33(8):1177–1187. doi:10.1016/j.jas.2005.12.006.

Zomer RJ, Trabucco A, Bossio DA, Verchot LV. 2008. Climate change mitigation: A spatial

analysis of global land suitability for clean development mechanism afforestation and

reforestation. Agriculture, Ecosystems & Environment. 126(1–2):67–80.

doi:10.1016/j.agee.2008.01.014.

109

APPENDICES

Appendix A: MSL 742 Mobility Maps

110

111

112

113

114

115

116

117

118

119

120

Appendix B: Geologic Age Table

Age (Ma)

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

VITA

MATTHEW HARRINGTON

Education: M.S. Geosciences, East Tennessee State University, Johnson

City, Tennessee, 2021

B.A. Geology & Environmental Studies, Augustana College,

Rock Island, Illinois, 2019

Public Schools, Glen Ellyn, Illinois

Professional Experience: Graduate Assistant, East Tennessee State University, College of

Arts and Sciences, 2019 – 2021

Teaching Assistant (Physical Geology of the Rocky Mountains)

Augustana College, 2018

Teaching Assistant (Dinosaurs and Extinction) Augustana College

2016 – 2019

Publications: Harrington, Matthew (2019) “Identifying Dietary and Migratory

Patterns of Illinois Woolly Mammoth Populations Using

Isotope Analysis of Carbon, Oxygen, and Strontium”

https://digitalcommons.augustana.edu/celebrationoflearning

/2019/posters/17

Honors and Awards: Best Paleontology Graduate Student, 2021

Best New Paleontology Graduate Student 2020

Best Undergraduate Research Poster, Triple Sectional GSA

Meeting, 2016

139