Late Quaternary Paleoecology of the North American Midcontinent

Home , Brickellia eupatorioides, Bromus latiglumis, Paleoecology

Iowa State University Capstones, Theses and Graduate Theses and Dissertations Dissertations

2020

Late Quaternary paleoecology of the North American midcontinent

Hannah M. Carroll Iowa State University

Follow this and additional works at: https://lib.dr.iastate.edu/etd

Recommended Citation Carroll, Hannah M., "Late Quaternary paleoecology of the North American midcontinent" (2020). Graduate Theses and Dissertations. 17846. https://lib.dr.iastate.edu/etd/17846

This Thesis is brought to you for free and open access by the Iowa State University Capstones, Theses and Dissertations at Iowa State University Digital Repository. It has been accepted for inclusion in Graduate Theses and Dissertations by an authorized administrator of Iowa State University Digital Repository. For more information, please contact [email protected]. Late Quaternary paleoecology of the North American midcontinent

Hannah M. Carroll

A dissertation submitted to the graduate faculty

in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

Major: Ecology, Evolution, and Organismal Biology; Environmental Science

Program of Study Committee: Lynn G. Clark, Co-major Professor Alan D. Wanamaker, Co-major Professor Beth E. Caissie Brian J. Wilsey Haldre S. Rogers

The student author, whose presentation of the scholarship herein was approved by the program of study committee, is solely responsible for the content of this dissertation. The Graduate College will ensure this dissertation is globally accessible and will not permit alterations after a degree is conferred.

Iowa State University

Ames, Iowa

2020

DEDICATION

This dissertation is for my grandmother, Linda Kurtz (1942 – 2018). She was, and remains, my hero and my most precious example of strength, intelligence, humor, and judiciously-applied stubbornness. My greatest wish was to defend this dissertation in her presence. Although she is not physically with me, her influence is carried with me in everything I do. I would not be me without her.

iii

TABLE OF CONTENTS

Page

ACKNOWLEDGMENTS ...... vi

ABSTRACT ...... viii

CHAPTER 1. GENERAL INTRODUCTION ...... 1 Dissertation objectives ...... 8 Dissertation organization ...... 9 Chapter 2: Ragweed and sagebrush pollen can distinguish between vegetation types at broad spatial scales ...... 9 Chapter 3: GeoPollen: An interactive, user-friendly Shiny Dashboard application for visualization and analysis of pollen data ...... 10 Chapter 4: Late Quaternary vegetation, fire, and hydroclimate at the southern limit of the temperate tallgrass prairie, Missouri, USA ...... 12 Other work ...... 13 References ...... 13

CHAPTER 2. RAGWEED AND SAGEBRUSH POLLEN CAN DISTINGUISH BETWEEN VEGETATION TYPES AT BROAD SPATIAL SCALES ...... 19 Abstract ...... 19 Introduction ...... 20 Methods ...... 24 Data acquisition and handling ...... 24 Ecoregion and climate space information ...... 27 Statistical analyses ...... 27 Results ...... 27 Pre-European Settlement ...... 30 Model Performance: Pre-European Settlement Precipitation Estimates ...... 31 Ecoregions ...... 32 Discussion ...... 35 Ecoregions ...... 37 Pre-European Settlement ...... 37 Conclusions ...... 38 Acknowledgements ...... 39 Author contributions...... 39 Data accessibility statement ...... 40 References ...... 40 Supporting Information ...... 43

CHAPTER 3. GEOPOLLEN: AN INTERACTIVE, USER-FRIENDLY SHINY DASHBOARD APPLICATION FOR VISUALIZATION AND ANALYSIS OF POLLEN DATA ...... 56 Abstract ...... 56 Introduction ...... 56

Materials and methods ...... 58 Data pre-processing ...... 58 Discussion of the GeoPollen interface ...... 59 The Main tab ...... 60 Modern Analog Technique calculations...... 61 Analog Communities tab ...... 62 Distribution Mapping tab ...... 64 Raw Data tab ...... 66 Other tabs ...... 66 Conclusions ...... 66 Acknowledgements ...... 67 Author contributions ...... 67 Data accessibility ...... 68 References ...... 68

CHAPTER 4. LATE QUATERNARY VEGETATION, FIRE, AND HYDROCLIMATE AT THE SOUTHERN LIMIT OF THE TEMPERATE TALLGRASS PRAIRIE, MISSOURI, USA ...... 70 Abstract ...... 70 Introduction ...... 71 Methods ...... 75 Study site ...... 75 Soil core extraction and processing ...... 76 Radiocarbon dating and age model creation ...... 77 Phytolith analysis ...... 78 Macroscopic charcoal analysis ...... 83 Stable carbon isotope methods ...... 83 X-ray diffraction ...... 85 Results ...... 85 Age model ...... 85 Phytoliths ...... 86 Other biogenic silica ...... 87 Phytolith indices ...... 87 Charcoal...... 88 Soil organic carbon ...... 89 Manganese oxide nodules ...... 90 Discussion ...... 94 Data availability ...... 99 References ...... 99 Acknowledgements ...... 105

CHAPTER 5. GENERAL CONCLUSIONS ...... 107 References ...... 108

APPENDIX A. GEOPOLLEN SOURCE CODE AND DOCUMENTATION ...... 110 Global.R ...... 110 Ui.R ...... 117

Server.R ...... 128 Get_downloads.R ...... 149 Userguide.Rmd ...... 150 Citeneotoma.Rmd ...... 160

APPENDIX B. PHYTOLITHS OF THE FLORA OF MISSOURI: A REFERENCE COLLECTION ...... 163 References ...... 166

APPENDIX C. USING LIGHT STABLE ISOTOPES TO ASSESS STREAM FOOD WEB ECOLOGY IN A GENERAL ECOLOGY LABORATORY COURSE ...... 167 Abstract ...... 167 Introduction ...... 168 Student Learning Outcomes (SLOs) ...... 173 Procedures ...... 174 Site Selection ...... 174 Field Sampling ...... 175 Results ...... 183 Sampling and Stable Isotopes...... 183 Data Analyses ...... 184 SLO Results ...... 184 Discussion ...... 189 Stable Isotopes Results ...... 189 Assessment ...... 191 Educational Impact ...... 192 Acknowledgements ...... 193 Disclosure Statement ...... 193 Data Accessibility Statement ...... 193 References ...... 194

ACKNOWLEDGMENTS

I would like to thank my PhD advisors, Dr. Lynn Clark and Dr. Alan Wanamaker, for welcoming me into their labs and supporting me through so many twists and turns over the years.

Lynn’s incredible strength and calm confidence have been an example I have needed over and over. Al’s warmth and encouragement has given me the courage to push past many hurdles and learn to trust myself. I have been incredibly fortunate to be guided by two such remarkable people, both personally and professionally.

I am deeply grateful to my committee, Dr. Richard Baker, Dr. Brian Wilsey, Dr. Haldre

Rogers, and Dr. Beth Caissie for their unfailing encouragement and support, and willingness to give me a gentle push in the right direction. Beth in particular has been an incredibly important source of strength, support, humor, and advice throughout my graduate career.

I would like to thank my friends at Iowa State University and elsewhere for endless support, laughter, coffee, and distractions. I could not have gotten through this without Karri

Folks, Dr. Rebecca Polich, Casey Judge, Grace Vaziri, Katie Thompson, Emily Ernst, Justin

Conover, and all of my longtime friends in the Nerd Cave. I am grateful to my friends and colleagues in the Clark and Wanamaker labs: Dr. Lakshmi Attigala, Dr. Tim Gallaher, Dr.

Thales Leandro, Monica Cox, Phil Klahs, Elizabeth McMurchie, Suzanne Ankerstjerne, Dr.

Maddie Mette, Diana Thatcher, Dr. Nina Whitney, and Juan Carlos Romero Galvez.

I am grateful to Dr. Ed Cushing for training me in pollen analysis and for being infinitely patient and kind. I wish to thank my undergraduate mentor at University of Washington,

Tacoma, Dr. Bonnie Becker, for giving me the footsteps in which to follow.

I am deeply grateful to my family for their unwavering support of every crazy thing I’ve decided to try over the years, getting a PhD being the craziest by far. My parents, Joe and Deana,

vii taught me that the path to a better life was education. I took them at their word, and this dissertation is the result. My dear brothers, Gabriel and William, give me a reason for being and keep me grounded. Their spouses, Julia and Torsten, are every bit as dear to me. My uncle Alex is the best uncle anyone could wish for and I want nothing more than to make him proud.

Finally, my nephew, Terrance, is the light of my life, and I hope to one day teach him all that

I’ve learned.

I am forever grateful to my boyfriend, Derek Houston, for supporting me through so very much over our time together. He has nursed me through an almost 2-year injury, brought me countless meals while I pushed through classes, prelims, research, and writing, walked miles upon miles with me when I was stressed, driven back and forth across the country for me more times than I can count, celebrated my smallest achievements, and even transcribed a large portion of my fourth chapter from my handwritten scrawl when prolonged eye strain left me unable to do anything except sit in the dark and write on paper. I am also grateful to Derek’s daughter, Tori, and sons, Logan and Mason, for their humor and kindness. I am grateful to

Derek’s parents, siblings, and their families for their support.

Finally, I would like to acknowledge that the land upon which Iowa State University is built is the ancestral homeland of the Báxoǰe (Ioway) people.

viii

ABSTRACT

Vegetation dynamics at regional to subcontinental scales are complex, and our understanding of the critical factors which govern them is far from complete. Temperature operates on a roughly longitudinal gradient across North America, while precipitation gradients are roughly latitudinal. What results is a complex climate space which is then further subdivided by topography, underlying geology, surface and subsurface hydrology, and seasonality, to name but a few, into broad-scale vegetation zones, or ecoregions. I used a combination of approaches to investigate climate and vegetation dynamics at multiple spatiotemporal scales, and to develop new tools to study past climate.

Defining and delineating past ecoregions has long presented a challenge. Fossil pollen distinguishes between major biome types, i.e., grassland versus forest, with great success.

However, distinguishing between prairie types at regional to subcontinental scales using the pollen record has been impossible. The ratio of Ambrosia (ragweed) to Artemisia (sagebrush, wormwood, mugwort) pollen has been shown to differentiate between tallgrass, mixed grass, and shortgrass prairie over a small portion of the modern Great Plains of North America. I extended that technique, and showed that the log-transformed Ambrosia to Artemisia ratio can reliably distinguish between subregions within the Great Plains, as well as regions immediately adjacent.

In addition, I found that the relationship between Ambrosia and Artemisia pollen is best explained by precipitation rather than temperature, and that it produces reliable precipitation estimates when used to create models. This will allow for better reconstructions of past climate and improve delineations of past ecoregion boundaries.

Pollen data are routinely used in paleoenvironmental studies to understand past climate and vegetation. One existing limitation in working with pollen data is the need to write code in

ix order to execute several of the routine analyses in paleoecological work. I developed GeoPollen, a Shiny Dashboard application, to be a streamlined, user-friendly GUI-based tool for performing these basic analyses. Users are able to utilize more than 3,000 publicly available pollen datasets from the Neotoma Paleoecology Database spanning the last 22,000 years from the United States and Canada. GeoPollen performs a suite of common tasks on demand and generates diagnostics necessary for evaluating results. I developed GeoPollen in order to increase the openness and accessibility of late Quaternary pollen data.

Boundaries between vegetation types are often highly sensitive to perturbations in climate. For example, the tallgrass prairie-temperate forest ecotone in Minnesota shifted rapidly and repeatedly during the mid-Holocene climate optimum, a warm and dry period ~9,000 - 5,000 years ago. A much less well-understood region exists at the southern limit of the prairie in southwest Missouri. This location serves as an excellent model system, as it represents a transitional zone between temperate grasslands and warm forests. I employed a multiproxy approach using soil carbon isotopes (a proxy for the C3:C4 ratio of the vegetation, and thereby hydroclimate), charcoal (a paleofire proxy), and phytoliths (taxonomically significant silica bodies produced by many plants) to examine vegetation dynamics at the tallgrass prairie-Ozark forest ecotone. Forest has not significantly invaded the prairie during the last 20,000 years, but the prairie has shifted from mixed grass (dry-mesic) to tallgrass (mesic) within the last 4,000 years. Fire likely played only a minor role until the mid-Holocene, after which it became frequent. This body of work serves to expand our understanding of the signals produced by vegetation in response to changing climate across the North American midcontinent throughout the late Quaternary.

CHAPTER 1. GENERAL INTRODUCTION

Patterns of vegetation distribution at regional to subcontinental scales are the result of a complex suite of interactions between biotic and abiotic factors working over multiple spatiotemporal scales. These factors range in scale from orbital forcing of the climate system, in which changes in the shape of Earth’s orbit around the sun and the angle of Earth’s axis affect the amount of solar radiation received by the Earth (i.e., Kutzbach et al. 1996, Harrison et al.

1998, Williams et al. 2001), to founder effects, i.e., the loss of genetic variation due to a small initial founding population, within the plant communities themselves, (Alsos et al. 2015).

Understanding the drivers of vegetation distribution has been a major focus of paleoecology since Gleason (1922) outlined the five floristic types present at the margins of the late-stage Wisconsin ice sheet. The roots of paleoecology, however, are much deeper. Several ancient Greek scholars working in the 7th through 5th Centuries BCE, possibly influenced by earlier Egyptian scholars, recognized fossils as the remains of once-living organisms (Ladd

1957). Some even attempted to connect them to the origins of animals and plants.

The concept of paleoecology ultimately traces its origins to the earliest-known paleoclimate reconstruction by the eminent Chinese polymath Shen Kuo, who lived in the 11th

Century CE. Kuo’s remarkable contribution was to recognize the link between long-term climatic changes and fossil evidence from bamboo (Needham 1959). He inferred that the presence of bamboo fossils found in Shaanxi Province, China, where they no longer grew, must point to past climatic conditions conducive to their survival. It would be several centuries before scholars would again recognize the link between fossils and past climate.

The next seminal work in the development of the broader field was Ristoro d’Arezzo’s

“Della Composizione del Mondo” (On the Composition of the World), published in 1282 CE. In

2 it, d’Arezzo explained the existence of fossil deposits by means of the diluvian hypothesis, then in vogue (Romano 2018a). D’Arezzo and other leading 13th Century Italian scholars were among the first to attempt to use actualism to explain the geologic record (Romano 2018a). It is noteworthy that, while Hutton and Lyell generally get the credit for originating the principle of uniformitarianism, the true origin is probably with these early scholars of actualism (Shea 1982).

Actualism was promoted again in the Renaissance by Leonardo da Vinci, a staunch antediluvian thinker. He recognized sediment transport processes in streams that lead to well- sorted marine deposits, and reasoned that fossils were laid down stratigraphically along with sediments (Romano 2018b). Da Vinci also explicitly noted that the internal structures of fossil shells bore a remarkable resemblance to their modern-day counterparts (Romano 2018a). He interpreted this similarity to mean that the environment had been relatively stable for these organisms; this stood in direct opposition to the diluvian notion of a catastrophic global flood.

Most significantly, da Vinci recognized that taphonomic processes could be inferred from bivalve shells. He discerned that fossil bivalve shells found separated were evidence of transport processing taking place following the death of the organism, and that whole shells were indicative of live burial (Mather and Mason 1939 in Cloud 1959). However, the notion that fossils were the preserved remains of plants and animals was far from universally accepted. The debate over the geologic processes which produced fossil deposits, and whether fossils had once been living organisms, raged throughout the Renaissance and well into the 18th Century (Cloud

1959).

The next major advance toward the establishment of paleoecology as a field of study arguably came in 1759, with the publication of Jean Étienne Guettard’s “On the accidents that have befallen fossil shells compared to those which are found to happen to shells now living in

3 the sea.” In it, Guettard argued that the same processes by which living marine organisms were battered and broken had applied long ago to those discovered as fossils (Fenton and Fenton

1952). His main intent was to convince skeptics, once and for all, that fossils had indeed been living organisms (Fenton and Fenton 1952, Cloud 1959). Guettard’s work had the dual effect of providing strong evidence that living assemblages could be used to contextualize fossils (Cloud

1959). By the late 18th Century, developments in the natural sciences started to occur rapidly.

Linnaeus was a contemporary of Guettard, and his establishment of the field of taxonomy revolutionized virtually every aspect of the biological sciences. Working just a few decades later, geologist James Hutton re-emphasized the existing principle of uniformitarianism (he is often incorrectly given credit for being its originator) (Shea 1982), and he was closely followed by

Lyell.

Georges Cuvier and Jean-Baptiste Lamarck are recognized as the fathers of paleontology, establishing the discipline around the turn of the 19th Century (Ladd 1957). They were far from alone in their field, however. Extensive fossil collection work went on throughout this period.

Most notably, Mary Anning and Elizabeth Philpot, who were contemporaries and friends, collected, described, and identified vast numbers of fossils. The fossils were essential to the development of paleontology, and thus paleoecology. Their work was admired and supported by

Henry de la Beche, a prominent geologist and founder of the Geological Survey of Great Britain, among other notables of the era (Torrens 1995).

The field of paleoecology as its own discipline emerged simultaneously with marine ecology with M. Edward Forbes’ landmark publication "Report on the Mollusca and Radiata of the Aegean Sea” in 1844 (Cloud 1959). Ecology itself would not become a formalized discipline for nearly 50 years, but the principles were firmly established by this time. The field of ecology

4 was proposed as a formal discipline by MIT chemist Ellen Swallow in 1892, although she was essentially ignored (Langenheim 1996), as was the case with most women scientists until the mid-20th Century. A man, J.B. Sanderson, co-opted her ideas and got credit for establishing the field instead (Sanderson 1893).

Meanwhile, important discoveries of the tools on which modern paleoecology depends were taking place. Pollen structure and function were extensively studied by a vast number of botanists beginning in the 17th Century (Ducker and Knox 1985), and it was this work that paved the way for the establishment of palynology in the early 20th Century. Phytoliths were discovered by Christian Ehrenberg in 1835 (Piperno 2006). Charles Darwin had unknowingly found phytoliths among the windborne microfossils collected aboard the H.M.S. Beagle in 1833.

He later sent dust samples to Ehrenberg, who identified and classified the phytoliths. Darwin discussed the “Phytolitharia” present in his dust samples and puzzled over their taphonomy in a paper published in the Quarterly Journal of the Geological Society (Darwin 1846). Another key tool, macrofossil analysis, the use of seeds, buds, twigs, and other macroscopic floral proxies, was developed in the 19th Century and continues to be an indispensable paleoecological proxy

(Birks 2001).

It was around the turn of the 20th Century when palynology, arguably the most commonly-used of the modern paleoecological tools, came into its own. Pollen analysis was born in the peatlands of Sweden in 1911, with the work of Gustaf Lagerheim and his protégé

Lennart von Post (Edwards 2018). Not long after, around 1918, von Post’s student Gunnar

Erdtman developed the first pollen diagram and proposed palynology as a tool for Quaternary paleoecology (Edlund and Winthrop 2014). Erdtman trained and worked with, for his time, a large number of women palynologists: at least 11 before the onset of World War II (Edwards

2018). Several of Erdtman’s students, women and men alike, would make substantial contributions to Quaternary paleoecology.

While Erdtman and colleagues were pushing the field of palynology forward in western

Europe, Phyllis Draper was doing the same in North America. Draper published the first pollen diagrams for North America (Draper 1929) and had a truly remarkable career, training more than a dozen Master’s students and one PhD student, many of whom were women. Her contributions to the field are inestimable.

The field of palynology exploded over the following decades and quickly became inextricably linked with paleoecology. The first major textbook on Quaternary pollen analysis was published by Gunnar Erdtman in 1943 (Erdtman 1943 in Birks and Berglund 2018). This was followed by “Text-Book of Pollen Analysis” by Fægri and Iversen (1950), and subsequent editions remain an important reference for the student of palynology. An extensive review of the history of pollen analysis can be found in Birks and Berglund (2018).

While palynology was growing, so too was phytolith analysis. Much of the research on phytoliths between the turn of the 20th Century and the start of World War II was conducted in

Germany (Piperno 1988). The application of phytoliths to ecological questions began in earnest in the 1950s, mostly in the UK, United States, and Australia, and continued through the mid-

1970s (Piperno 2006). The modern period of research followed in the late 1970s and continues to be highly active to the present day (Piperno 2006). Phytolith taxonomy and reporting have been standardized by the International Committee for Phytolith Taxonomy, who recently released their second version of the accepted nomenclature (International Committee for Phytolith

Taxonomy (ICPT) et al. 2019). The second edition nomenclature is used in this dissertation.

Modern Quaternary paleoecology in the United States was largely advanced by the

Minnesota school, headed by Herb Wright, Jr., whose work included a vast array of paleoecological topics. He collaborated internationally, visiting with dozens of top researchers in

Minnesota and abroad, including Hilary and John Birks, Magnus Fries, Bill Watts, Barbara

Hansen, Knut Fægri, and even Gunnar Erdtman himself (Birks 2017). Many of Wright’s students went on to make major advances in paleoecology, notably Ed Cushing, Dick Baker, Dan

Engstrom, Barbara Hansen, and many others (Birks 2017). Researchers trained by Wright have played a huge role in transforming Quaternary paleoecology from a young, qualitative discipline focused on localized site reconstructions to a fully mature quantitative field, working at regional to subcontinental scales.

Major advances toward quantitative research in paleoecology were made by Margaret

Davis, who was the first to consider the connection between the spatial distribution of pollen grains recovered in sediments and the population size of the vegetation which had produced it

(Davis 1963). This paper served to push the field toward the regional reconstructions that are now a hallmark of Quaternary paleoecology. She painstakingly documented the landscape-scale range shifts of trees throughout the late Quaternary over an illustrious career that spanned nearly four decades.

Quaternary paleoecology began moving toward landscape and regional scales in the

1970s through the 1990s with the publication of major syntheses by Baker (1983), Delcourt and

Delcourt (1983), Wright (1984), and Gaudreau (1988), among others, with a small number of authors pushing toward continental-scale syntheses (e.g., Livingstone 1975). One crucial development was the advent of the earliest pollen databases, including COHMAP, the

Cooperative Holocene Mapping Project in the 1970s and 80s, followed by the North American

Pollen Database in the early 1990s (for a detailed history of pollen database development, see

Grimm et al. 2018). The Modern Analog Technique (MAT) was introduced by Overpeck et al.

(1985) and remains one of the most powerful and commonly used analyses in late Quaternary paleoecology. These important efforts provided the toolkits essential to today’s paleoecologist.

Other seminal papers around this time employed ever-larger datasets to reconstruct major climatic shifts (e.g., Baker et al. 1992, Wright 1992) and quantify the complementarity of independent paleoproxies (e.g., Baker et al. 1998). Modeling work began to come to the fore

(e.g., Sugita et al. 1999), paving way for today’s big data applications.

Subcontinental to continental scale work became possible around 2000 CE thanks to improvements in computing power, and efforts to develop new statistical tools intensified.

Climate reconstruction and modeling at broad spatiotemporal scales developed relatively quickly

(e.g., Davis 2000, Davis and Shaw 2001, Gavin et al. 2003, Jackson and Williams 2004,

Williams and Jackson 2007, Nordt et al. 2008), much of which was supported by the pollen databases developed over the previous decades, and the Neotoma Paleoecology Database in 2007

(Williams et al. 2018).

At present, Quaternary paleoecology is being pushed forward by an intensely collaborative field of researchers working on everything from traditional site-level reconstructions to global modeling of paleoclimate and vegetation distribution. Phytolith analysis is a highly active field and has contributed much to our understanding of vegetation and climate.

Recent works which have spurred my fascination with the field of Quaternary paleoecology include Stromberg (2004), Gill et al. (2009), Bartlein et al. (2011), Brewer et al. (2012), Blois et al. (2013), Voelker et al. (2015), Shuman and Marsicek (2016), and especially the modeling of

Commerford et al. (2018).

While I was working to become proficient in the paleoecological literature, a major synthesis of the state of the field was published by Seddon et al. (2014). In it, Seddon and colleagues outline what they argue are 50 of the most pressing questions in paleoecology. This paper inspired and informed much of what became my dissertation research. In particular,

Question 39, “How can common environmental signals be identified in multiple records at different spatial and temporal scales?” sparked my imagination. I am intensely interested in the environmental signals buried within massive amounts of proxy data. The question posed by

Seddon et al. (2014) prompted me to think seriously about what I might be able to contribute.

My interest in history also led me to dive deeply into the literature surrounding the development of my discipline, from the earliest recognition of pollen as a potential paleoproxy, through the advent of big data tools made possible by high-powered computing. I learned how researchers initially began to approach problems of spatiotemporal scale, and how the quest to identify common signals lead to massive efforts to standardize, compile, synthesize, and report paleoecological data over the whole globe. A consideration of the history and development of my field helps to give context to the projects I pursue. My dissertation research comprises an effort to contribute two new tools for identifying common environmental signals at broad spatial scales, and to assist in improving spatiotemporal data coverage in understudied regions.

Dissertation objectives

The specific aim of my dissertation was to improve the identification of a common set of environmental signals at multiple spatiotemporal scales over the late Quaternary. My approach to this was as follows:

1) Evaluate the applicability of a pollen ratio derived from two common taxa, ragweed and sagebrush, to reconstruct past vegetation type and precipitation over relatively recent timescales across the entirety of the North American midcontinent.

2) Develop an application to aid the research community in conducting analyses and identifying patterns across all currently available pollen data in North America spanning the last

22,000 years.

3) Examine vegetation, fire, and hydroclimate history in an ecologically unique and understudied region of the North American midcontinent, and evaluate the timing of major climatic events in the late Quaternary in this region versus those in more extensively studied areas.

Dissertation organization

Chapter 2: Ragweed and sagebrush pollen can distinguish between vegetation types at broad spatial scales

Defining and delineating past ecoregions has long presented a substantial challenge to paleoecologists. Fossil pollen has traditionally been employed with great success to distinguish between major biome types, i.e., grassland versus forest. However, distinguishing between prairie types at regional to subcontinental scales using the pollen record has presented a seemingly insurmountable challenge. Standard light microscopy, in use by the vast majority of palynologists, does not allow for types of grass pollen to be distinguished from one another, and no other consistent markers of tallgrass versus mixed grass versus shortgrass prairie had been found that were appropriate over broad spatial scales.

The ratio of Ambrosia (ragweed) to Artemisia (sagebrush, wormwood, mugwort) pollen was shown by Morris (2013) to differentiate between tallgrass, mixed grass, and shortgrass prairie over a small portion of the climate space occupied by the modern Great Plains of North

America. Carroll et al. (in press) refined and extended that technique, and were able to show that the log transformed Ambrosia to Artemisia ratio can reliably distinguish between subregions within the Great Plains, as well as adjacent regions immediately bordering the Great Plains. We

10 then demonstrated that the relationship between Ambrosia and Artemisia pollen is driven by precipitation rather than temperature. We used the log Ambrosia to Artemisia ratio to model modern precipitation, and showed that the models could then be applied to produce Little Ice

Age (commonly from CE 1450 to 1850) precipitation estimates that agreed with independent, non-pollen proxies. In addition, we documented that the median proportion of Ambrosia pollen has more than doubled since European settlement of the midcontinent, and that Artemisia has been more or less unaffected. This provides evidence that a simple correction can be applied to post-settlement pollen records in order to filter out the settlement signal and align records with their presettlement counterparts. This will allow for better reconstructions of past climate and improve delineations of past ecoregion boundaries.

Chapter 3: GeoPollen: An interactive, user-friendly Shiny Dashboard application for visualization and analysis of pollen data

Pollen data are a standard tool used in paleoecological studies to reconstruct past climate and vegetation. Massive efforts to standardize and house data produced by thousands of researchers over several decades have led to the advent of the Neotoma Paleoecology Database

(Williams et al. 2018). This repository provides free and open access to data, and encourages the development of new tools. However, in order to work with large amounts of data, it is necessary to be proficient in writing R code. This presents a point-of-entry barrier in many cases. To address this barrier, I developed GeoPollen.

GeoPollen is a Shiny Dashboard application which provides a simple, user-friendly GUI interface to visualize, synthesize, and perform basic analyses on pollen data publicly available through Neotoma. As of March 2020, approximately 1,500 datasets from nearly 1,300 unique sites, translating to nearly 60,000 rows of data, are available for North America over the last

22,000 years. GeoPollen fully automates the data cleaning and processing necessary for dealing

11 with tens of thousands of lines of pollen data. All data cleaning (such as searching for and removing incomplete records and accidental duplications) and processing is done silently at app startup, and the user is only shown the final form unless they choose to download raw data. The clean data then feed into the rest of the app's functions. Users can search for data by age range and/or geographic location on the Main tab, and by age range and taxon on the Distribution tab.

Reactive tables constantly update to display metadata and basic stats as the user makes selections and moves around the map. There are built-in functions that allow users to specify a training set and test set and do modern analog calculations using the Squared Chord Distance metric. It provides diagnostic output and produces publication-quality heatmaps. These are functions that are currently impossible to access without a coding background. GeoPollen additionally provides the first GUI-based option for doing modern analog calculations using Squared Chord Distance

(currently the standard for pollen analysis) of which I am aware. This tool is meant to forward the Neotoma community’s mission of making paleo data open and accessible to the broadest possible audience.

GeoPollen is intended to be an ongoing project. Hosting is provided by the College of

Liberal Arts and Sciences at Iowa State University, and support is provided by the Research IT department. New data uploaded to Neotoma are incorporated weekly via an automated R script.

Quarterly app updates and improvements are expected to continue for the foreseeable future. The source code used for the release version of GeoPollen (Version 1.0) is given in Appendix A. For the latest version of GeoPollen’s source code, users may visit the GitHub repository: https://github.com/hannahcarroll/GeoPollen.

Chapter 4: Late Quaternary vegetation, fire, and hydroclimate at the southern limit of the temperate tallgrass prairie, Missouri, USA

Chapters 2 and 3 dealt exclusively with “big data” explorations of the paleoecological record. Part of the outcome of exploring those datasets at subcontinental scales is a greater awareness of regions which are understudied. One such region that stands out is the tallgrass prairie-Ozark forest ecotone. Although this region is ecologically and climatologically unique, the absence of natural lakes, necessary for pollen preservation, has hindered research. Few pollen records exist in this area, and those that do exist tend to be deep in the Ozark forest, far from the ecotone.

I identified a suitable study location approximately two miles from the present-day tallgrass prairie-Ozark forest ecotone (as defined by the Commission for Environmental

Cooperation and Secretariat (1997)) in southwest Missouri, a tallgrass prairie remnant known as

Golden Prairie. Golden Prairie is the largest and highest-quality remnant prairie in Missouri. It is owned and managed by the Missouri Prairie Foundation, who were incredibly supportive of having a paleoecological study done at this site. With their permission, I retrieved two intact soil cores containing the full soil profile of the site in November 2017. From those cores, I measured stable carbon and nitrogen isotopes in bulk soil as a proxy for the past C3:C4 ratio, and thereby hydroclimate; macroscopic charcoal as a proxy for regional fire activity; and phytoliths as a direct characterization of the past plant communities. I discovered a complex vegetational history with a highly C4-dominated soil organic carbon signal throughout the entirety of the last 19,000 years. I also found signals of continuously open habitat, highly unusual for the region. Fire has likely played a role in maintaining the prairie since the mid-Holocene. Much more work is needed to determine whether Golden Prairie is in and of itself unique in the region, or if vegetation signals here have been confounded by other factors. However, it is probable that

13 vegetation and climate at Golden Prairie over much of the late Quaternary has been reflective of a steep east-west moisture gradient, which is considerably weaker at present.

To support the phytolith portion of my paleoclimate reconstruction, I created a phytolith reference collection for the phytolith-bearing flora of Missouri. All reference material was obtained from the Ada Hayden Herbarium (ISC) at Iowa State University. Appendix B provides information on that collection and links to access the images in the international phytolith database known as PhytCore. In addition, a physical copy of my reference collection is permanently housed in the Ada Hayden Herbarium.

Other work

I developed a passion while in graduate school for experiential learning techniques in undergraduate science education. Experiential learning has been repeatedly shown to enhance the learning of all students, but to disproportionately benefit students from underrepresented backgrounds (e.g., Hurtado et al. 2010). Helping to remove barriers to entry in STEM is a major focus of my career aspirations. Appendix C consists of a manuscript published in Journal of

Biological Education in which Carroll et al. (2019) present a module for teaching stream food web ecology to undergraduates using experiential learning techniques. We show how to employ stable isotope analysis of macroinvertebrate and fish tissues to promote learning of major ecological concepts, practical field and laboratory methods, statistical analyses in R, and data interpretation.

References

Alsos, I. G., D. Ehrich, P. B. Eidesen, H. Solstad, K. B. Westergaard, P. Schönswetter, A. Tribsch, S. Birkeland, R. Elven, and C. Brochmann. 2015. Long-distance plant dispersal to North Atlantic islands: colonization routes and founder effect. AoB PLANTS 7:plv036.

Baker, R. G. 1983. Late Quaternary environments of the United States: The late Pleistocene. 433 pp. (H. E. Wright and S. C. Porter, Eds.). U of Minnesota Press.

Baker, R. G., L. A. Gonzalez, M. Raymo, E. A. Bettis, M. K. Reagan, and J. A. Dorale. 1998. Comparison of multiple proxy records of Holocene environments in the midwestern United States. Geology 26:1131–1134.

Baker, R. G., L. J. Maher, C. A. Chumbley, and K. L. Van Zant. 1992. Patterns of Holocene environmental change in the Midwestern United States. Quaternary Research 37:379– 389.

Bartlein, P. J., S. P. Harrison, S. Brewer, S. Connor, B. A. S. Davis, K. Gajewski, J. Guiot, T. I. Harrison-Prentice, A. Henderson, O. Peyron, I. C. Prentice, M. Scholze, H. Seppä, B. Shuman, S. Sugita, R. S. Thompson, A. E. Viau, J. Williams, and H. Wu. 2011. Pollen- based continental climate reconstructions at 6 and 21 ka: a global synthesis. Climate Dynamics 37:775–802.

Birks, H. H. 2001. Plant macrofossils. Pages 49–74 in J. P. Smol, H. J. B. Birks, W. M. Last, R. S. Bradley, and K. Alverson, editors. Tracking Environmental Change Using Lake Sediments: Terrestrial, Algal, and Siliceous Indicators. Springer Netherlands, Dordrecht.

Birks, H. J. B. 2017. Herbert E. Wright, Jr., Biographical memoir. National Academy of Sciences.

Birks, H. J. B., and B. E. Berglund. 2018. One hundred years of Quaternary pollen analysis 1916–2016. Vegetation History and Archaeobotany 27:271–309.

Blois, J. L., J. W. Williams, M. C. Fitzpatrick, S. Ferrier, S. D. Veloz, F. He, Z. Liu, G. Manion, and B. Otto-Bliesner. 2013. Modeling the climatic drivers of spatial patterns in vegetation composition since the Last Glacial Maximum. Ecography 36:460–473.

Brewer, S., S. T. Jackson, and J. W. Williams. 2012. Paleoecoinformatics: applying geohistorical data to ecological questions. Trends in Ecology & Evolution 27:104–112.

Carroll, H. M., D. D. Houston, S. Ankerstjerne, and A. D. W. Jr. 2019. Using light stable isotopes to assess stream food web ecology in a general ecology laboratory course. Journal of Biological Education.

Carroll, H. M., A. D. Wanamaker, L. G. Clark, and B. J. Wilsey. in press. Ragweed and sagebrush pollen can distinguish between vegetation types at broad spatial scales. Ecosphere.

Cloud, P. E. 1959. Paleoecology: retrospect and prospect. Journal of Paleontology 33:926–962.

Commerford, J. L., E. C. Grimm, C. J. Morris, A. Nurse, I. Stefanova, and K. K. McLauchlan. 2018. Regional variation in Holocene climate quantified from pollen in the Great Plains of North America. International Journal of Climatology 38:1794–1807.

Commission for Environmental Cooperation, and Secretariat. 1997. Ecological regions of North America: toward a common perspective. The Commission, Montréal, Québec.

Darwin, C. 1846. An account of the fine dust which often falls on vessels in the Atlantic Ocean. Quarterly Journal of the Geological Society 2:26–30.

Davis, M. B. 1963. On the theory of pollen analysis. American Journal of Science 261:897–912.

Davis, M. B. 2000. Palynology after Y2K—understanding the source area of pollen in sediments. Annual Review of Earth and Planetary Sciences 28:1–18.

Davis, M. B., and R. G. Shaw. 2001. Range shifts and adaptive responses to Quaternary climate change. Science 292:673–679.

Delcourt, P. A., and H. R. Delcourt. 1983. Late-Quaternary vegetational dynamics and community stability reconsidered. Quaternary Research 19:265–271.

Draper, P. 1929. A comparison of pollen spectra of old and young bogs in the Erie Basin. Proceedings of the Oklahoma Academy of Science 9:50–53.

Ducker, S. C., and R. B. Knox. 1985. Pollen and pollination: A historical review. Taxon 34:401– 419.

Edlund, A. F., and Z. A. Winthrop. 2014. Sharing what he saw: An appreciation of Gunnar Erdtman’s life and illustrations. Grana 53:1–21.

Edwards, K. J. 2018. Pollen, women, war and other things: reflections on the history of palynology. Vegetation History and Archaeobotany 27:319–335.

Erdtman, G. 1943. An introduction to pollen analysis. Chronica Botanica Company, Waltham, MA.

Fægri, K., and J. Iversen. 1950. Text-book of modern pollen analysis. Ejnar Munksgaard, Copenhagen, Denmark.

Fenton, C. L., and M. A. Fenton. 1952. Giants of geology. Doubleday & Company, Inc., Garden City, New York.

Gaudreau, D. C. 1988. The distribution of late Quaternary forest regions in the northeast. Pages 215–256 in G. P. Nicholas, editor. Holocene Human Ecology in Northeastern North America. Springer US, Boston, MA.

Gavin, D. G., W. W. Oswald, E. R. Wahl, and J. W. Williams. 2003. A statistical approach to evaluating distance metrics and analog assignments for pollen records. Quaternary Research 60:356–367.

Gill, J. L., J. W. Williams, S. T. Jackson, K. B. Lininger, and G. S. Robinson. 2009. Pleistocene megafaunal collapse, novel plant communities, and enhanced fire regimes in North America. Science 326:1100–1103.

Gleason, H. A. 1922. The vegetational history of the middle west. Annals of the Association of American Geographers 12:39–85.

Grimm, E. C., J. Blois, T. Giesecke, R. Graham, A. Smith, and J. Williams. 2018. Constituent databases and data stewards in the Neotoma Paleoecology Database: History, growth, and new directions. Past Global Change Magazine 26:64–65.

Harrison, S. P., D. Jolly, F. Laarif, A. Abe-Ouchi, B. Dong, K. Herterich, C. Hewitt, S. Joussaume, J. E. Kutzbach, J. Mitchell, N. D. Noblet, and P. Valdes. 1998. Intercomparison of simulated global vegetation distributions in response to 6 kyr BP orbital forcing. Journal of Climate 11:2721–2742.

Hurtado, S., C. B. Newman, M. C. Tran, and M. J. Chang. 2010. Improving the rate of success for underrepresented racial minorities in STEM fields: Insights from a national project. New Directions for Institutional Research 2010:5–15.

International Committee for Phytolith Taxonomy (ICPT), K. Neumann, C. A. E. Strömberg, T. Ball, R. M. Albert, L. Vrydaghs, and L. S. Cummings. 2019. International code for phytolith nomenclature (ICPN) 2.0. Annals of Botany 124:189–199.

Jackson, S. T., and J. W. Williams. 2004. Modern analogs in Quaternary paleoecology: Here today, gone yesterday, gone tomorrow? Annual Review of Earth and Planetary Sciences 32:495–537.

Kutzbach, J., G. Bonan, J. Foley, and S. P. Harrison. 1996. Vegetation and soil feedbacks on the response of the African monsoon to orbital forcing in the early to middle Holocene. Nature 384:623–626.

Ladd, H. S. 1957. Treatise on marine ecology and paleoecology 2: Paleoecology. Geological Society of America.

Langenheim, J. H. 1996. Early history and progress of women ecologists: Emphasis upon research contributions. Annual Review of Ecology and Systematics 27:1–53.

Livingstone, D. A. 1975. Late Quaternary climatic change in Africa. Annual Review of Ecology and Systematics 6:249–280.

Mather, K. F., and S. L. Mason. 1939. A source book in geology. First edition. McGraw-Hill, London.

Morris, C. J. 2013. Analysis of modern pollen data from the prairies of central North America. Thesis, Kansas State University, Manhattan, Kansas.

Needham, J. 1959. Science and civilisation in China Vol. 3. Cambridge University Press, Cambridge, England.

Nordt, L., J. Von Fischer, L. Tieszen, and J. Tubbs. 2008. Coherent changes in relative C4 plant productivity and climate during the late Quaternary in the North American Great Plains. Quaternary Science Reviews 27:1600–1611.

Overpeck, J. T., T. Webb, and I. C. Prentice. 1985. Quantitative interpretation of fossil pollen spectra: Dissimilarity coefficients and the method of modern analogs. Quaternary research. 23:87–108.

Piperno, D. R. 1988. Phytolith analysis: An archaeological and geological perspective. Academic Press, San Diego, CA.

Piperno, D. R. 2006. Phytoliths: a comprehensive guide for archaeologists and paleoecologists. Rowman Altamira.

Romano, M. 2018a. Palaeoecology before ecology: the rise of actualism, palaeoenvironment studies and palaeoclimatology in the Italian panorama between the fourteenth and eighteenth centuries. Italian Journal of Geosciences 137:16–30.

Romano, M. 2018b. Italian Diluvianism and antidiluvianism within the international arena: the great debate that lasted more than six centuries. Proceedings of the Geologists’ Association 129:17–39.

Sanderson, J. B. 1893. On the origin of biology and its relation to the other branches of natural science. British Medical Journal 2:613–617.

Seddon, A. W. R., A. W. Mackay, A. G. Baker, H. J. B. Birks, E. Breman, C. E. Buck, E. C. Ellis, C. A. Froyd, J. L. Gill, L. Gillson, E. A. Johnson, V. J. Jones, S. Juggins, M. Macias-Fauria, K. Mills, J. L. Morris, D. Nogués-Bravo, S. W. Punyasena, T. P. Roland, A. J. Tanentzap, K. J. Willis, M. Aberhan, E. N. van Asperen, W. E. N. Austin, R. W. Battarbee, S. Bhagwat, C. L. Belanger, K. D. Bennett, H. H. Birks, C. Bronk Ramsey, S. J. Brooks, M. de Bruyn, P. G. Butler, F. M. Chambers, S. J. Clarke, A. L. Davies, J. A. Dearing, T. H. G. Ezard, A. Feurdean, R. J. Flower, P. Gell, S. Hausmann, E. J. Hogan, M. J. Hopkins, E. S. Jeffers, A. A. Korhola, R. Marchant, T. Kiefer, M. Lamentowicz, I. Larocque-Tobler, L. López-Merino, L. H. Liow, S. McGowan, J. H. Miller, E. Montoya, O. Morton, S. Nogué, C. Onoufriou, L. P. Boush, F. Rodriguez-Sanchez, N. L. Rose, C. D. Sayer, H. E. Shaw, R. Payne, G. Simpson, K. Sohar, N. J. Whitehouse, J. W. Williams, and A. Witkowski. 2014. Looking forward through the past: identification of 50 priority research questions in palaeoecology. Journal of Ecology 102:256–267.

Shea, J. H. 1982. Twelve fallacies of uniformitarianism. Geology 10:455–460.

Shuman, B. N., and J. Marsicek. 2016. The structure of Holocene climate change in mid-latitude North America. Quaternary Science Reviews 141:38–51.

Strömberg, C. A. E. 2004. Using phytolith assemblages to reconstruct the origin and spread of grass-dominated habitats in the great plains of North America during the late Eocene to early Miocene. Palaeogeography, Palaeoclimatology, Palaeoecology 207:239–275.

Sugita, S., M.-J. Gaillard, and A. Broström. 1999. Landscape openness and pollen records: a simulation approach. The Holocene 9:409–421.

Torrens, H. 1995. Presidential Address: Mary Anning (1799-1847) of Lyme; “The greatest fossilist the world ever knew.” The British Journal for the History of Science 28:257– 284.

Voelker, S. L., M. C. Stambaugh, R. P. Guyette, X. Feng, D. A. Grimley, S. W. Leavitt, I. Panyushkina, E. C. Grimm, J. P. Marsicek, B. Shuman, and B. Brandon Curry. 2015. Deglacial hydroclimate of midcontinental North America. Quaternary Research 83:336– 344.

Williams, J. W., E. C. Grimm, J. L. Blois, D. F. Charles, E. B. Davis, S. J. Goring, R. W. Graham, A. J. Smith, M. Anderson, J. Arroyo-Cabrales, A. C. Ashworth, J. L. Betancourt, B. W. Bills, R. K. Booth, P. I. Buckland, B. B. Curry, T. Giesecke, S. T. Jackson, C. Latorre, J. Nichols, T. Purdum, R. E. Roth, M. Stryker, and H. Takahara. 2018. The Neotoma Paleoecology Database, a multiproxy, international, community- curated data resource. Quaternary Research 89:156–177.

Williams, J. W., and S. T. Jackson. 2007. Novel climates, no-analog communities, and ecological surprises. Frontiers in Ecology and the Environment 5:475–482.

Williams, J. W., B. N. Shuman, and T. Webb. 2001. Dissimilarity analyses of late-Quaternary vegetation and climate in eastern North America. Ecology 82:3346–3362.

Wright, H. E. 1984. Sensitivity and response time of natural systems to climatic change in the late quaternary. Quaternary Science Reviews 3:91–131.

Wright, H. E. 1992. Patterns of Holocene climatic change in the midwestern United States. Quaternary Research 38:129–134.

CHAPTER 2. RAGWEED AND SAGEBRUSH POLLEN CAN DISTINGUISH BETWEEN VEGETATION TYPES AT BROAD SPATIAL SCALES

Hannah M. Carroll,1 Alan D. Wanamaker,2 Lynn G. Clark,1 and Brian J. Wilsey1

1Department of Ecology, Evolution, and Organismal Biology, Iowa State University, Ames,

Iowa 50011 USA; 2Department of Geological and Atmospheric Sciences, Iowa State University,

Ames, Iowa 50011 USA

Modified from a manuscript in press in Ecosphere

Abstract

Patterns of vegetation distribution at regional to subcontinental scales can inform understanding of climate. Delineating ecoregion boundaries over geologic time is complicated by the difficulty of distinguishing between prairie types at broad spatial scales using the pollen record. Pollen ratios are sometimes employed to distinguish between vegetation types, although their applicability is often limited to a geographic range. The Neotoma Paleoecology Database offers an unparalleled opportunity to synthesize a large number of pollen datasets. Ambrosia

(ragweed) is a genus of mesic-adapted species sensitive to summer moisture. Artemisia

(sagebrush, wormwood, mugwort) is a genus of dry-mesic adapted species resilient to drought.

The log pollen ratio between these two common taxa was calculated across the North American midcontinent from surface pollen samples housed in the Neotoma Paleoecology Database. The relative proportion of Ambrosia has roughly doubled since European settlement, likely due to widespread disturbance, while Artemisia proportions are nearly unchanged. Correcting surface samples for the disturbance signal in modern Ambrosia proportions will allow Ambrosia, a strong indicator of summer moisture, to be more accurately represented. In surface samples where both Ambrosia and Artemisia are reported as nonzero proportions of the pollen sum, mean annual precipitation explains approximately 78% of the variation in the log Ambrosia to

Artemisia ratio. Application of this model to Little Ice Age pollen samples produces precipitation reconstructions which generally agree with reconstructions from independent non-pollen proxies.

In addition, we find that modern ecoregions within the North American midcontinent can be successfully distinguished from one another using the log Ambrosia to Artemisia ratio. These relationships can improve reconstructions of past climate and improve delineation of past ecoregion boundaries.

Introduction

Vegetation distributions at regional to subcontinental scales are a result of the interaction between climate, fire, dispersal, disturbance, and competition over millennia (Williams 2008).

These complex processes create distinct regions of habitat that change dramatically over geologic time in response to both external and internal climate forcings. Understanding past vegetation distributions has been a major focus of paleoecological work since Gleason (1922) outlined the five floristic types present at the margins of the late-stage Wisconsin ice sheet.

Subsequent research refined and extended Gleason’s groundbreaking ideas from localized to landscape scales (Baker et al. 1992), from landscape to regional scales (Davis and Shaw 2001), and from regional to subcontinental scales (Williams et al. 2004). Delineating past ecoregion boundaries has long presented a challenge due to the need to access, standardize, and synthesize large amounts of data through space and time.

Reconstructions of past ecoregion distributions have improved dramatically with the advent of paleoecological databases such as the North American Pollen Database (NAPD),

FAUNal MAP (FAUNMAP), and the Neotoma Paleoecology Database, the latter of which now serves as the primary repository for pollen data in the western hemisphere, and has partially subsumed the two former (Grimm et al. 2018). These massive efforts provide access to thousands of contributions from individual researchers and make it possible to stitch records

21 together over landscape and subcontinental scales, and from glacial to modern time periods. It is now possible to refine estimates of past ecoregion extent, and therefore climate, at ever-higher spatiotemporal resolutions.

Grasslands have dominated much of the North American midcontinent since the early

Miocene (Axelrod 1985, Edwards et al. 2010). Current vegetation assemblages, however, are post-glacial in origin (Jackson and Williams 2004). Prairies quickly expanded their range as the climate warmed following the Last Glacial Maximum, and reclaimed much of the North

American midcontinent by the mid-Holocene (Wright 1992, Baker et al. 2000). The modern tallgrass prairie region extends from the Blackland Prairie region of east-central Texas, USA, to southwestern Saskatchewan, Canada, and east into the prairie peninsula of Illinois (Risser et al.

1981) (Figure 2-1). The western boundary of the contemporaneous tallgrass prairie region extends from east-central Texas through eastern North Dakota before veering westward into far west Saskatchewan. Through all but the most northerly portion, the western boundary is flanked by regions of mixed grass and shortgrass prairie extending from southeastern Saskatchewan to

New Mexico and into southern Texas. An international, multi-agency effort to precisely delineate the present-day extent of all major vegetation types across the continent resulted in the publication of the Ecoregions of North America (Commission for Environmental Cooperation

1997).

Figure 2-1: Map of the EPA Level II Ecoregions of North America as defined by CEC (1997). Thick black outline is the boundary of the Great Plains region. Thick blue outline is the 2 decimal degree buffer used to select pollen datasets from the Neotoma Paleoecology Database.

One existing limitation in the effort to improve ecoregion delineations over geologic time is the difficulty of distinguishing between prairie types at broad spatial scales using the pollen record. Multiple approaches have been proposed with mixed levels of success. In the southwestern United States, an eight-taxon pollen signature derived from the modern pollen rain was shown to successfully differentiate between extant vegetation types (Hoyt 2000), but this study covered a small geographical area with extreme differences in climate unlike those experienced by much of the Great Plains region. The ratio of Artemisia to Chenopodiaceae pollen has been applied to differentiate between vegetation types in arid to semi-arid regions worldwide, but was shown to be applicable only within a narrow precipitation range of 450 to

500 mm annually, and is not comparable across geographic regions (Zhao et al. 2012). The ratio of Artemisia to Cyperaceae pollen was shown to differentiate reliably between high-alpine meadow and steppe, but is likely only applicable in these regions (Herzschuh 2007).

One promising technique for delineating ecoregions is the ratio of Ambrosia to Artemisia, which has been shown to differentiate between tallgrass, mixed grass, and shortgrass prairie over a small portion of the climate space occupied by the modern Great Plains of North America

(Morris 2013). Both Ambrosia and Artemisia are wind-pollinated dicotyledons in the Asteraceae

(sunflower family), subfamily Asteroideae. Ambrosia (Supertribe Helianthodae, Tribe

Heliantheae, Subtribe Ambrosiinae) (Baldwin 2009), commonly known as ragweed, is a genus of mesic-adapted, herbaceous plants notable for high disturbance tolerance. Artemisia (Supertribe

Asterodae, Tribe Anthemideae) (Oberprieler et al. 2009), variously called sagebrush, wormwood, and mugwort, is a genus of dry-adapted woody or herbaceous plants. Ambrosia, in general, is sensitive to summer moisture, while Artemisia is tolerant of drought, except for some winter sensitivity (Grimm 2001). Although the ranges of Ambrosia and Artemisia overlap over most of the North American continent, their autecology differs enough to employ them as complementary indicators. It is this feature of their habitat tolerances that makes them particularly useful as a tool to aid in delineating ecoregions.

We here extend the Ambrosia to Artemisia ratio technique of Morris (2013) to the full geographical extent of the modern Great Plains, as defined by CEC (1997), and demonstrate its ability to distinguish between adjacent ecoregions, as well as prairie types, over broad geographical areas.

A noteworthy complication present in the modern pollen record is the extreme disturbance created by the arrival of large numbers of European colonists in the 19th century.

This disturbance appears in the pollen record as a sharp increase in the proportion of Ambrosia pollen that has been dubbed the “settlement horizon.” Pollen assemblages before and after this horizon cannot typically be treated as analogs, even if they are closely separated in time and

24 space (Kujawa et al. 2016). For this reason, many studies necessarily exclude the highly disturbed Midwestern United States from their modern analog training sets. We present a tentative correction for this disturbance signal so that records from this region may be incorporated into training sets.

In this study, we evaluate the effectiveness of using the Ambrosia to Artemisia ratio as a technique to robustly delineate ecoregions in modern presettlement pollen records. We hypothesize that the ratio of Ambrosia to Artemisia pollen can be used to distinguish between ecoregions of the North American midcontinent. Additionally, we outline a statistical approach to better align presettlement assemblages with their modern counterparts in order to assess how pollen assemblages change through space and time. We hypothesize that, following appropriate correction for post-settlement disturbance, the ratio of Ambrosia to Artemisia pollen can be used to reconstruct past precipitation.

Methods

Data acquisition and handling

Pollen data were obtained from the Neotoma Paleoecology Database (Williams et al.

2018) via package ‘neotoma’ ver. 1.7.2. (Goring et al. 2015). All analyses were performed in R ver. 3.5.1 (R Core Team 2018). A full list of datasets included in this study is given in

Supporting Information (Tables S 2-1:S 2-2 [surface pollen records] and S 2-3:S 2-4

[presettlement pollen records]). All source code used for this project is available at https://github.com/hannahcarroll/GeoPollen.

Pollen records for the North American midcontinent were downloaded from Neotoma as type “pollen,” (i.e., fossil pollen), or type “surface sample,” (i.e., modern pollen). Taxa from raw assemblages were standardized using the Whitmore Full list (Whitmore et al. 2005). Non-pollen palynomorphs and spikes (i.e., microspheres or Lycopodium tablets) were removed from the

25 dataset before calculations were made. Iva pollen is sometimes indistinguishable from that of

Ambrosia, and therefore we cannot exclude the possibility that Iva contributes to reported

Ambrosia counts in some regions.

Raw age models for fossil pollen datasets were calibrated to IntCal09 after (Reimer et al.

2009). Fossil pollen datasets were then subset to just those that were 250 to 500 years in age, corresponding to Little Ice Age climate in the North American midcontinent before the well- documented and sharp increase in the proportion of Ambrosia pollen resulting from the arrival of

European colonists and implementation of widespread land clearing, known as the “settlement horizon”. Hereafter, these datasets are referred to as “presettlement.”

Pollen records were clipped to within two decimal degrees (approximately 222 km) of the boundary of the Great Plains Region (Figure 2-1) as defined by (Commission for Environmental

Cooperation 1997). A two decimal degree buffer was chosen because it is approximately double the 50% source area of ragweed pollen entering the center of a basin 750 m in diameter as predicted by (Sugita 1993). Basin size is unknown for most of the datasets obtained from

Neotoma, and therefore source distance can only be roughly estimated. The two decimal degree buffer around the boundary of the Great Plains region accounts for the source area of pollen entering all but the largest of basins and reflects the uncertainty inherent in delineating ecoregion boundaries. We excluded datasets in both surface and presettlement sets that were located within regions which were likely receiving most or all of their pollen rain from non-grassland taxa.

These comprise the modern-day Western Cordillera, Cold Deserts, and Upper Gila Mountains.

A total of 549 records (462 unique sites) of “surface sample” type (hereafter, “surface”) were returned by Neotoma. Of these, 26 were located in the Western Cordillera region and 5 were located in the Cold Deserts, and were removed. Of those that remained (518), 11 records

(2.1%) reported neither Ambrosia nor Artemisia and were removed. 507 records were included in analyses. 83 records (16.4%) included Ambrosia but no Artemisia. 101 records (19.9%) did not include Ambrosia but did include Artemisia. Both taxa were reported as nonzero in 323 records

(63.7%). Records in which at least one of the two taxa were reported as nonzero were retained.

A total of 843 records (211 unique sites), each spanning a portion of the past 250-500 calendar years before present, were available from Neotoma at the time of analysis. We removed

62 records which were located within the 2-degree buffer region but at high elevations, and likely to reflect pollen from non-grassland taxa. These were sites in the modern-day Western

Cordillera (58 datasets), Cold Deserts (3), and Upper Gila Mountains (1). Of those that remained,

68 (8.7%) contained neither Ambrosia nor Artemisia and were removed. The remaining 713 records (169 unique sites) were included in analyses. Of these, 13 records (1.8%) reported

Ambrosia but no Artemisia, and 64 (8.98%) reported Artemisia but no Ambrosia. Both Ambrosia and Artemisia were reported in 636 datasets (89.2%).

Locations in which either Ambrosia or Artemisia, but not both, were reported represent ecologically distinct areas and are important to include in analyses. Therefore, in both surface and presettlement records in which either taxon was reported as zero, and the other nonzero, the proportion was recoded to a dummy value. This allows variation in the relative proportion of the nonzero taxon to be expressed in the resulting ratio. Zero occurrences of either Ambrosia or

Artemisia were recoded to 1x10-5. This coding produces a ratio two orders of magnitude different than the smallest ratio in either dataset in which both Ambrosia and Artemisia were nonzero. The ratio of Ambrosia to Artemisia is calculated from the relative proportion of each in their respective records. In surface records, the Ambrosia proportion was divided by the median

27 increase relative to the presettlement proportion (2.201) prior to the calculation of the ratio. The log ratio was used in all analyses.

Ecoregion and climate space information

Site location information for each record was used to match records to their respective

EPA Level II ecoregions via spatial join in R package ‘sp’ (Pebesma and Bivand 2005). Records were then joined to a raster containing climate information from the Bioclim 2 dataset (Fick and

Hijmans 2017) at a 2.5 arcminute (roughly 5 km) resolution. For a table of the number of datasets per ecoregion included in analyses, see Table S 2-5.

Statistical analyses

Discriminant function analysis was used to test whether ecoregion could successfully be predicted from the Ambrosia to Artemisia ratio and precipitation in the surface pollen dataset.

Data were z-scored (centered and scaled) using the package ‘caret’ (Kuhn et al. 2019).

Discriminant function analysis was performed in package ‘mda’ (S original by Hastie et al.

2017) on a split training set (80%) and a test set (20%) bootstrapped in package ‘boot’ (Davison and Hinkley 1997, Canty and Ripley 2019) using 1,000 random splits. Ecoregions with fewer than 10 datasets were excluded (Softwood Shield and Texas-Louisiana Coastal Plain).

Results

The proportion of Ambrosia pollen in the surface pollen dataset (median proportion

0.059) is approximately double (2.201) that of the presettlement set (median proportion 0.027).

Only small changes were found between the Artemisia proportion in surface (median proportion

0.02) versus presettlement samples (median proportion 0.031) (Figure 2-2).

Figure 2-2: Relative proportion of Ambrosia and Artemisia reported in presettlement versus surface pollen records across all datasets included in analyses.

After correcting the surface dataset for the doubling in median proportion between presettlement and surface Ambrosia, where both taxa were present, the log Ambrosia to Artemisia ratio ranged from -5.61 to 4.22 with a mean of 0.13 (Table 2-1; Figure 2-3). In cases where zero Artemisia was reported, the log Ambrosia to Artemisia ratio ranged from 4.65 to 10.16 with a mean of 8.35.

In cases where zero Ambrosia was reported, the log Ambrosia to Artemisia ratio ranged from -

11.83 to -7.17 with a mean of -9.36.

Table 2-1: Summaries of the surface pollen Ambrosia to Artemisia ratio by presence or absence of Ambrosia and Artemisia. Group n Mean Log Ratio Min Log Ratio Max Log Ratio Both Present 323 0.13 -5.61 4.22 Zero Ambrosia 101 -9.36 -11.83 -7.17 Zero Artemisia 83 8.35 4.65 10.16

Figure 2-3: Log Ambrosia to Artemisia ratio between scenarios in the surface set. For records in which either taxon was reported as zero, and the other nonzero, the proportion was recoded to a dummy value of 1x10-5. This allows variation in the relative proportion of the nonzero taxon to be expressed in the resulting ratio. 507 records are included in analyses. 323 records report both Ambrosia and Artemisia, 101 records report Artemisia but no Ambrosia, and 83 records report Ambrosia but no Artemisia.

Co-occurrences of Ambrosia and Artemisia were reported across a wide range of precipitation in the surface set, from 269 to 1,182 mm annually, with a mean of 598 mm. Where

Ambrosia occurred but Artemisia did not, precipitation ranged from 492 to 1,631 mm annually, with a mean of 1,006 mm. Where Artemisia occurred but Ambrosia did not, precipitation ranged from 295 to 579 mm annually, with a mean of 458 mm.

The relationship between mean annual precipitation (MAP) and the log Ambrosia to

Artemisia ratio differs among the three scenarios. Where either Ambrosia or Artemisia, but not both, occurs, the relationship between MAP and the log ratio is mostly linear (Figure 2-4). MAP for the zero Ambrosia cases explains 31.0% of the variation in the Ambrosia to Artemisia ratio.

Figure 2-4: The relationship between precipitation and the log Ambrosia to Artemisia ratio for the surface pollen dataset is linear for the zero Artemisia and zero Ambrosia cases, and exponential where both are present.

The model is weak for the zero Artemisia cases, with MAP explaining 21.7% of the variation in the log Ambrosia to Artemisia ratio. Where both types of pollen are present, the relationship between MAP and the Ambrosia to Artemisia ratio is modeled using a bootstrapped B-spline fit with three knots (Figure S 2-1), which explains 77.7% of the variation in the Ambrosia to

Artemisia ratio and produces normally distributed residuals (Figure S 2-2). The relationship between MAP and the Ambrosia to Artemisia ratio is positive between approximately 200 mm and 800 mm MAP. Between approximately 800 mm and the maximum of approximately 1,200 mm MAP, the slope of the relationship between the log Ambrosia to Artemisia ratio and precipitation is positive but decreases slightly.

Pre-European Settlement

The surface set models were applied to a presettlement dataset in order to validate the relationships between the Ambrosia to Artemisia ratio and precipitation estimates. Where both taxa were present (706 records), the log Ambrosia to Artemisia ratio ranged from -5.37 to 4.04,

31 with a mean of -0.28. Where Artemisia was present but Ambrosia was not (75 records), the log

Ambrosia to Artemisia ratio ranged from -10.69 to -5.34, with a mean of -7.59.

Model Performance: Pre-European Settlement Precipitation Estimates

Application of the B-spline model to presettlement Ambrosia to Artemisia log ratios produced a mean MAP of 538.5 mm (range 349.4 mm to 927.3 mm) where both Ambrosia and

Artemisia are present in the record (Table 2-2). This is, on average, 135.8 mm/year less than modern observed precipitation over the midcontinent (range -209.0 mm/year to 677.9 mm/year less relative to modern values; Figure 2-5).

Table 2-2: Precipitation values for the presettlement dataset. The zero Artemisia model has a low slope, producing a very high likelihood of erroneous precipitation values and is therefore not recommended for use. Difference (mm) Group Modern MAP (mm) Reconstructed MAP (mm) (Modern – Reconstructed) Mean Min Max Mean Min Max Mean Min Max Both present 674.2 346.7 1178.2 538.5 349.4 927.3 135.8 -209.0 677.9 Zero Ambrosia 498.5 357.9 823.1 525.0 408.8 609.2 -26.5 -93.8 266.3 Zero Artemisia 970.8 720.4 1178.2 1079.6 920.2 1183.4 -108.8 -391.4 171.8

Figure 2-5: Difference in modern versus reconstructed precipitation at all presettlement dataset locations. Diagonal line is 1:1.

Where Artemisia is present but Ambrosia is not, the linear model produced a mean annual precipitation of 525.0 mm and resulted in a reconstructed presettlement precipitation at these locations, on average, 26.5 mm/year greater than at present (Figure 2-6).

Figure 2-6: Difference in modern versus reconstructed precipitation at presettlement sites. Blue indicates reconstructed LIA precipitation is higher than at present, yellow indicates modern and LIA precipitation that is roughly similar, and red indicates drier conditions than at present.

There are too few cases of zero Artemisia in the presettlement set (13 records from 7 unique locations) to adequately test the model. In the surface pollen set, the model fit to zero Artemisia cases is not only weak, but also produces a low slope, leading to a high risk of error. It is unlikely to produce realistic precipitation estimates and we therefore exclude it from further consideration.

Ecoregions

The Ambrosia to Artemisia ratio varies across the midcontinent, being generally low in the northern regions, and increasing toward the south and east (Figure 2-7). The ratio is closely

33 tied to precipitation (Figure 2-8a, b), and varies less with temperature (Figure 2-8c). See

Supporting Information Figure S 2-3 for an interactive 3-dimensional visualization.

Figure 2-7: Map of the log Ambrosia to Artemisia ratio for every surface sample in the analysis.

Discriminant analysis (DA) suggests that an additive model including the log Ambrosia to

Artemisia ratio and MAP can distinguish between ecoregions with between 48.5% and 77.3% success, producing a mean success rate of 61.3% on 1,000 random partitions of the dataset.

Figure 2-8: a) Three-dimensional plot of the log Ambrosia to Artemisia ratio versus mean annual precipitation and mean annual temperature by EPA Level II Ecoregion. b) Log Ambrosia to Artemisia ratio versus mean annual precipitation and c) mean annual temperature by EPA Level II Ecoregion. For an interactive 3-dimensional version of this figure, see Supporting Information Figure S 2-3.

ANOVA followed by adjustment for multiple comparisons using Tukey’s Honest Significant

Difference (HSD) indicates that there are statistically significant differences in the Ambrosia to

Artemisia ratio between the three subregions of the Great Plains: West-Central Semiarid Prairies vs Temperate Prairies (Difference = -2.13, p = 0.002), Temperate Prairies vs. South-Central

Semiarid Prairies (Difference = -3.09, p < 0.0001), and West-Central Semiarid Prairies vs. South

Central Semiarid Prairies (Difference = -5.22, p < 1x10-7) (Figure 2-9; See Table S 2-6 for a full pairwise comparison between ecoregions).

Figure 2-9: The log Ambrosia to Artemisia ratio for all surface pollen samples by ecoregion.

Inclusion of the interaction between latitude and longitude (to control for spatial autocorrelation) produced no change in the ANOVA coefficients and was therefore excluded from the final model.

Discussion

Vegetation responses to climate are complex, yet predictable patterns often emerge. The

Ambrosia to Artemisia ratio is strongly driven by precipitation in a predictable pattern across the

North American midcontinent. Ambrosia species in this region tend to be moderately drought sensitive, especially in summer, whereas Artemisia demonstrates high drought tolerance, with some sensitivity in winter (Grimm et al. 2011). These differing climate tolerances make it possible to use their relationship to each other to understand hydroclimate at multiple spatiotemporal scales. Where both taxa are present, MAP explains nearly 78% of the variation in the Ambrosia to Artemisia ratio, and the addition of temperature as a variable does not markedly improve the model. Where no Artemisia is present and Ambrosia is reported as nonzero, the model is extremely weak and its use is inadvisable. This may be due to differences in moisture sensitivity; Artemisia does not occur in the surface pollen set where moisture is above 1200 mm annually, whereas Ambrosia occurs over the full range of precipitation in the dataset. It may also be a result of spatial bias in surface pollen samples housed in the Neotoma Paleoecology

Database, which has been thoroughly analyzed by Inman et al. (2018). In the zero Artemisia cases, the linear relationship is strongly affected by groups of samples taken within a very narrow range of precipitation and the removal of those groups effectively changes the apparent relationship with precipitation. For this reason, the Ambrosia to Artemisia ratio should not be applied where Artemisia is absent. Records where Ambrosia is absent and Artemisia is present invariably occur in areas that receive less than ~600 mm of precipitation annually; given that the median relative proportion of Artemisia is only slightly different in pre-European settlement versus modern records, this appears to be an especially strong aridity signal.

Our findings agree with (Commerford et al. 2018), who report that inclusion of Ambrosia in transfer functions drove precipitation estimates, and that precipitation appears to be a much stronger factor in Ambrosia pollen abundance than is temperature. We suggest that the strong signal of disturbance in the post-European settlement pollen record can be adequately controlled

37 for using a correction factor. Our simple two-taxon ratio also avoids the difficulties inherent in accounting for post-settlement disturbance to a large number of taxa with a wide variety of apparent disturbance signals.

Ecoregions

(Morris 2013) successfully employed the Ambrosia to Artemisia ratio to distinguish between tallgrass, mixed grass, and shortgrass prairie over a small region within the Great Plains.

Our models suggest the tool is widely applicable over the entirety of the Great Plains region, and that modern ecoregions carry distinct signals driven largely by precipitation. This can improve our ability to define past vegetational boundaries and refine precipitation estimates at a range of spatial scales. We find tentative support for our hypothesis that the ratio of Ambrosia to

Artemisia pollen can be used to distinguish between ecoregions of the North American midcontinent.

Pre-European Settlement

We reconstruct unchanged to slightly increased LIA precipitation in the central Great

Plains relative to the present, with areas of moderately to substantially reduced precipitation at the eastern and western margins of the Great Plains. A paucity of presettlement pollen records in the central Great Plains, because of the scarcity of natural lakes, hampers precise estimation.

However, tentative conclusions can be drawn from the relatively few records that are available, and we here consider independent precipitation proxies (salinity, lake level, stable isotope, tree ring, speleothem, etc.) to evaluate the ability of the Ambrosia to Artemisia ratio to accurately reconstruct precipitation.

Reconstructions of reduced precipitation north-central Great Plains immediately preceding European settlement broadly agree with records based on lake levels and salinity in eastern North Dakota (Fritz et al. 1994, Valero-Garcés et al. 1997). We reconstruct unchanged to

38 slightly wetter conditions over the same period in south-central Canada, which is corroborated by reduced lake salinity near Medicine Hat, Alberta (Vance et al. 1992). High apparent spatial heterogeneity of moisture near the northern and northwestern limits of the tallgrass prairie region has been reported from diatom-inferred salinity records and may indicate instability of the jet stream during the LIA (Laird et al. 2003).

Strong drought signals in the midwestern United States, especially Minnesota,

Wisconsin, and Illinois, must be interpreted with care. Precipitation in the upper Midwest, USA, during the LIA was increased, and fire activity reduced, relative to the MCA (Shuman and

Marsicek 2016). However, precipitation since the LIA has increased markedly. We find tentative support for our hypothesis that, following appropriate correction for post-settlement disturbance, the ratio of Ambrosia to Artemisia pollen can be used to reconstruct past precipitation.

Conclusions

We provide past precipitation estimates to demonstrate that our calibration of modern pollen to modern precipitation produces models which can be applied across time and space.

Precipitation reconstructions broadly agree with the majority of independent proxy records available throughout the North American midcontinent during the LIA. Despite pollen being a primarily regional signal, high spatial heterogeneity of precipitation appears to be reflected in the records satisfactorily. The ratio is strongly driven by precipitation, while temperature appears to be a less critical factor. It remains to be seen whether this holds true further back in time when climate was much different than at present. It is also unknown whether the Ambrosia to

Artemisia ratio is applicable outside of the North American midcontinent. Most pollen ratio methods are applicable only to the regions for which they are developed, although the Ambrosia to Artemisia ratio is apparently useful over a much larger area than most. However, Ambrosia is

39 an introduced species outside of North America and the use of the ratio on other continents is not advisable.

Our reconstructions of past precipitation generally agree with independent non-pollen proxies. Limitations occur where spatiotemporal coverage of pollen datasets is poor. However, for well-studied areas, the application of a correction factor to the modern Ambrosia proportion may enable researchers to include it in training sets and precipitation models. It remains to be seen whether alternative plant-derived markers (macrofossils, phytoliths, etc.) may be employed in a similar manner.

Acknowledgements

The authors thank K.A. Moloney and G. Kuttubekova for valuable suggestions regarding data analysis. We thank H.S. Rogers, R.G. Baker, E.A. Ernst, D.D. Houston, and two anonymous reviewers for comments on earlier versions of the manuscript that greatly improved its quality.

Data were obtained from the Neotoma Paleoecology Database (http://www.neotomadb.org) and its constituent database, the North American Pollen Database. The work of data contributors, data stewards, and the Neotoma community is gratefully acknowledged. No outside funding supported this work.

Author contributions

HMC wrote the R code used in this work, performed analyses, and wrote the manuscript.

BJW contributed the idea for the project and provided guidance on analyses and interpretation.

ADW and LGC provided further guidance in the development of the manuscript. All authors read the manuscript and contributed to its development. The authors declare that they have no conflicts of interest.

Data accessibility statement

All data used in analyses are publicly available through the Neotoma Paleoecology

Database (http://www.neotomadb.org). Raw data obtained from Neotoma and R scripts used to perform analyses are available via a public GitHub repository at https://github.com/hannahcarroll/AmbrosiatoArtemisia.

References

Axelrod, D. I. 1985. Rise of the grassland biome, central North America. The Botanical Review 51:163–201.

Baker, R. G., G. G. Fredlund, R. D. Mandel, and E. A. I. Bettis. 2000. Holocene environments of the central Great Plains: multi-proxy evidence from alluvial sequences, southeastern Nebraska. Quaternary International 67:75–88.

Baker, R., L. Maher, C. Chumbley, and K. Van Zant. 1992. Patterns of Holocene environmental change in the midwestern United States. Quaternary Research 37:379–389.

Baldwin, B. G. 2009. Heliantheae alliance. Pages 689–707 in V. A. Funk, A. Susanna, T. F. Stuessy, and R. J. Bayer, editors. Systematics, evolution, and biogeography of Compositae. International Association for Plant Taxonomy, Vienna, Austria.

Canty, A., and B. Ripley. 2019. boot: Bootstrap R (S-Plus) Functions.

Commission for Environmental Cooperation. 1997. Ecological regions of North America: Toward a common perspective. (map update) Commission for Environmental Cooperation, Montreal.

Davis, M. B., and R. G. Shaw. 2001. Range shifts and adaptive responses to Quaternary climate change. Science 292:673–9.

Davison, A. C., and D. V. Hinkley. 1997. Bootstrap methods and their applications. Cambridge University Press, Cambridge.

Edwards, E. J., C. P. Osborne, C. A. E. Stromberg, S. A. Smith, W. J. Bond, P. A. Christin, A. B. Cousins, M. R. Duvall, D. L. Fox, R. P. Freckleton, O. Ghannoum, J. Hartwell, Y. Huang, C. M. Janis, J. E. Keeley, E. A. Kellogg, A. K. Knapp, A. D. B. Leakey, D. M. Nelson, J. M. Saarela, R. F. Sage, O. E. Sala, N. Salamin, C. J. Still, and B. Tipple. 2010. The origins of C4 grasslands: Integrating evolutionary and ecosystem science. Science 328:587–591.

Fick, S. E., and R. J. Hijmans. 2017. WorldClim 2: New 1‐km spatial resolution climate surfaces for global land areas. International Journal of Climatology 37:4302–4315.

Fritz, S. C., D. R. Engstrom, and B. J. Haskell. 1994. “Little Ice Age” aridity in the North American Great Plains: a high-resolution reconstruction of salinity fluctuations from Devils Lake, North Dakota, USA. The Holocene 4:69–73.

Gleason, H. A. 1922. The vegetational history of the middle west. Annals of the Association of American Geographers 12:39–85.

Goring, S., A. Dawson, G. L. Simpson, K. Ram, R. W. Graham, E. C. Grimm, and J. W. Williams. 2015. neotoma: A programmatic interface to the Neotoma Paleoecological Database. Open Quaternary 1:1–17.

Grimm, E. C. 2001. Trends and palaeoecological problems in the vegetation and climate history of the northern Great Plains , U.S.A. Biology and Environment: Proceedings of the Royal Irish Academy 101B:47–64.

Grimm, E. C., J. L. Blois, T. Giesecke, R. W. Graham, A. J. Smith, and J. W. Williams. 2018. Constituent databases and data stewards in the Neotoma Paleoecology Database : History, growth, and new directions. PAGES Magazine 26:64–65.

Grimm, E. C., J. J. Donovan, and K. J. Brown. 2011. A high-resolution record of climate variability and landscape response from Kettle Lake, northern Great Plains, North America. Quaternary Science Reviews 30:2626–2650.

Hastie, T., R. Tibshirani (S original), F. Original R port by Leisch, K. Hornik, and B. D. Ripley. 2017. mda: Mixture and Flexible Discriminant Analysis.

Herzschuh, U. 2007. Reliability of pollen ratios for environmental reconstructions on the Tibetan Plateau. Journal of Biogeography 34:1265–1273.

Hoyt, C. A. 2000. Pollen signatures of the arid to humid grasslands of North America. Journal of Biogeography 27:687–696.

Inman, R., J. Franklin, T. Esque, and K. Nussear. 2018. Spatial sampling bias in the Neotoma paleoecological archives affects species paleo-distribution models. Quaternary Science Reviews, 198:115-125.

Kuhn, M., J. Contributions from Wing, S. Weston, A. Williams, C. Keefer, A. Engelhardt, T. Cooper, Z. Mayer, B. Kenkel, the R. C. Team, M. Benesty, R. Lescarbeau, A. Ziem, L. Scrucca, Y. Tang, C. Candan, and T. Hunt. 2019. caret: Classification and Regression Training.

Kujawa, E. R., S. Goring, A. Dawson, R. Calcote, E. C. Grimm, S. C. Hotchkiss, S. T. Jackson, E. A. Lynch, J. McLachlan, J. M. St-Jacques, C. Umbanhowar, and J. W. Williams. 2016. The effects of anthropogenic land cover change on pollen-vegetation relationships in the American Midwest. Anthropocene 15:60–71.

Laird, K. R., B. F. Cumming, S. Wunsam, J. A. Rusak, R. J. Oglesby, S. C. Fritz, and P. R. Leavitt. 2003. Lake sediments record large-scale shifts in moisture regimes across the northern prairies of North America during the past two millennia. Proceedings of the National Academy of Sciences of the United States of America 100:2483–2488.

Morris, C. J. 2013. Ananalysis of modern pollen data from the prairies of central North America. Kansas State University, Manhattan, Kansas.

Oberprieler, C., S. Himmelreich, M. Källersjö, J. Vallès, L. E. Watson, and R. Vogt. 2009. Anthemideae. Pages 631–666 in V. A. Funk, A. Susanna, T. F. Stuessy, and R. J. Bayer, editors. Systematics, evolution, and biogeography of Compositae. International Association for Plant Taxonomy, Vienna, Austria.

Pebesma, E. J., and R. S. Bivand. 2005. Classes and methods for spatial data in R. R News 5:9– 13.

R Core Team. 2018. R: A language and environment for statistical computing. Foundation for Statistical Computing, Vienna, Austria.

Reimer, P. J., M. G. L. Baillie, E. Bard, A. Bayliss, J. W. Beck, P. G. Blackwell, C. B. Ramsey, C. E. Buck, G. S. Burr, R. L. Edwards, M. Friedrich, P. M. Grootes, T. P. Guilderson, I. Hajdas, T. J. Heaton, A. G. Hogg, K. A. Hughen, K. F. Kaiser, B. Kromer, F. G. McCormac, S. W. Manning, R. W. Reimer, D. A. Richards, J. R. Southon, S. Talamo, C. S. M. Turney, J. van der Plicht, and C. E. Weyhenmeyer. 2009. IntCal09 and Marine09 radiocarbon age calibration curves, 0-50,000 years CAL BP. Radiocarbon 51:1111–1150.

Risser, P. G., E. C. Birney, H. D. Blocker, S. W. May, W. J. Parton, J. A. Wiens, and Editors. 1981. The true prairie ecosystem (US/IBP synthesis series, vol. 16). Hutchinson Ross, Stroudsburg, Pennsylvania, USA.

Shuman, B. N., and J. Marsicek. 2016. The structure of Holocene climate change in mid-latitude North America. Quaternary Science Reviews 141:38–51.

Sugita, S. 1993. A model of pollen source area for an entire lake surface. Quaternary Research 39:239–244.

Valero-Garcés, B. L., K. R. Laird, S. C. Fritz, K. Kelts, E. Ito, and E. C. Grimm. 1997. Holocene Climate in the Northern Great Plains Inferred from Sediment Stratigraphy, Stable Isotopes, Carbonate Geochemistry, Diatoms, and Pollen at Moon Lake, North Dakota. Quaternary Research 48:359–369.

Vance, R. E., R. W. Matthewes, and J. J. Clague. 1992. 7000 year record of lake-level change on the northern Great Plains: a high-resolution proxy of past climate. Geology 20:879–882.

Whitmore, J., K. Gajewski, M. Sawada, J. W. Williams, B. Shuman, P. J. Bartlein, T. Minckley, A. E. Viau, T. Webb, S. Shafer, P. Anderson, and L. Brubaker. 2005. Modern pollen data from North America and Greenland for multi-scale paleoenvironmental applications. Quaternary Science Reviews 24:1828–1848.

Williams, J. W. 2008. Quaternary vegetation distributions. Pages 1–15 in V. Gornitz, editor. Encyclopedia of Paleoclimatology and Ancient Environments. Springer Science & Business Media.

Williams, J. W., B. N. Shuman, T. Webb, P. J. Bartlein, and P. L. Leduc. 2004. Late-Quaternary vegetation dynamics in north America: Scaling from taxa to biomes.

Wright, H. E. 1992. Patterns of holocene climatic change in the midwestern United States. Quaternary Research 38:129–134.

Zhao, Y., H. Liu, F. Li, X. Huang, J. Sun, W. Zhao, U. Herzschuh, and Y. Tang. 2012. Application and limitations of the Artemisia/Chenopodiaceae pollen ratio in arid and semi-arid China. The Holocene 22:1385–1392.

Supporting Information

Table S 2-1: Citation information for all surface pollen records included in this study for which publication information is available.

Dataset Citation ID MacDonald, G.M. 1984. Postglacial plant migration and vegetation development in the western 190 Canadian boreal forest. Dissertation. University of Toronto, Toronto, Ontario, Canada. Baker, R.G., L.J. Maher, Jr., C.A. Chumbley, and K.L. Van Zant. 1992. Patterns of Holocene 202 environmental change in the midwestern United States. Quaternary Research 37(3):379-389. [DOI: 10.1016/0033-5894(92)90074-S] Watts, W.A., and H.E. Wright, Jr. 1966. Late-Wisconsin pollen and seed analysis from the Nebraska 207 Sandhills. Ecology 47:202-210. McAndrews, J.H., and H.E. Wright, Jr. 1969. Modern pollen rain across the Wyoming basins and the 257 northern Great Plains (U.S.A.). Review of Palaeobotany and Palynology 9:17-43. McAndrews, J.H. 1966. Postglacial history of prairie, savanna, and forest in northwestern Minnesota. 265 Torrey Botanical Club Memoir 22(2):1-72. Swain, P.C. 1979. The development of some bogs in eastern Minnesota. Doctoral dissertation. 402 University of Minnesota, Minneapolis, Minnesota, USA. Webb, T., III. 1970. The late- and post-glacial sequence of climatic events in Wisconsin and east- 453 central Minnesota: quantitative estimates derived from fossil pollen spectra by multivariate statistical analysis. Dissertation. University of Wisconsin, Madison, Wisconsin, USA. Peterson, G.M. 1978. Pollen spectra from surface sediments of lakes and ponds in Kentucky, Illinois, 651 and Missouri. American Midland Naturalist 100:333-340. Kapp, R.O. 1965. Illinoian and Sangamon vegetation in southwestern Kansas and adjacent 680 Oklahoma. Contributions from the Museum of Paleontology 19:167-255.

Table S 2-1 (continued)

Dataset Citation ID Mott, R.J. 1969. Palynological studies in Central Saskatchewan: Contemporary pollen spectra from 682 surface samples. Geological Survey of Canada Paper 69-32. Lichti-Federovich, S., and J.C. Ritchie. 1968. Recent pollen assemblages from the western interior of 685 Canada. Review of Palaeobotany and Palynology 7(4):297-344. [DOI: 10.1016/0034- 6667(68)90035-3] Delcourt, P.A., H.R. Delcourt, and J.L. Davidson. 1983. Mapping and calibration of modern pollen- 689 vegetation relationships in the southeastern United States. Review of Palaeobotany and Palynology 39:1-45. Van Zant, K.L., T. Webb III, G.M. Peterson, and R.G. Baker. 1979. Increased Cannabis/Humulus 740 pollen, an indicator of European agriculture in Iowa. Palynology 3:227-234. Hall, S.A. 1977. Holocene geology and paleoenvironmental history of the Hominy Creek Valley. 743 Pages 12-42 in D.O. Henry, editor. The Prehistory and paleoenvironment of Hominy Creek Valley. Laboratory of Archeology, University of Tulsa, Tulsa, Oklahoma, USA. Hall, S.A. 1977. Geology and palynology of archeological sites and associated sediments. Pages 13- 744 41 in D.O. Henry, editor. The prehistory of the Little Caney River, 1976 field season. Laboratory of Archeology, University of Tulsa, Tulsa, Oklahoma, USA. Peters, M.A., and T. Webb III. 1979. A radiocarbon dated pollen diagram from west-central 745 Wisconsin. Bulletin of the Ecological Society of America 60:102. Davis, A.M. 1975. Reconstructions of local and regional Holocene environments from the pollen and 764 peat stratigraphies of some Driftless Area peat deposits. Doctoral dissertation. University of Wisconsin, Madison, Wisconsin, USA. Lichti-Federovich, S., and J. C. Ritchie. 1965. Contemporary pollen spectra in central Canada. II. 3700 The forest-grassland transition in Manitoba. Pollen et Spores 7:63-87. Commerford, J.L., K.K. McLauchlan, and S. Sugita. 2013. Calibrating vegetation cover and 7464 grassland pollen assemblages in the Flint Hills of Kansas, USA. American Journal of Plant Sciences 4(7A):1-10. [DOI: 10.4236/ajps.2013.47A1001] Parker, N.E., and J.W. Williams. 2011. The influence of climate, cattle density, and lake morphology on Sporomiella abundances in modern lake sediments in the US Great Plains. The Holocene 8368 22(4):475-483. http://hol.sagepub.com/content/early/2011/11/28/0959683611425550. [DOI: 10.1177/0959683611425550] Commerford, J.L., E.C. Grimm, C.J. Morris, A. Nurse, I. Stefanova, and K.K. McLauchlan. 2018. 11478 Regional variation in Holocene climate quantified from pollen in the Great Plains of North America. International Journal of Climatology 38(4):1794-1807. [DOI: 10.1002/joc.5296]

Table S 2-2: Dataset ID and contact name for all surface pollen datasets for which full citation information is not available.

Contact ID Contact Name 11 Baker, Richard G. 28 Baker, Richard G. 32 Davis, Anthony M. 33 Delcourt, Hazel R. 44 Delcourt, Paul A. 53 Grimm, Eric Christopher 63 Kapp, Rold O. 67 Lichti-Federovich, Sigrid 69 Maher, Louis J., Jr. 74 McAndrews, John H. 82 Mott, Robert J.

Table S 2-2 (continued)

Contact ID Contact Name 86 Peters, M. A. 99 Ritchie, James C. 101 Swain, Patricia C. 101 Van Zant, Kent L. 106 Van Zant, Kent L. 108 Watts, William A. 108 Webb, Thompson III 114 Webb, Thompson III 125 Wright, Herbert E., Jr. 136 Chumbley, Craig A. 146 Hall, Stephen A. 163 MacDonald, Glen 163 Peterson, G. M. 685 Peterson, G. M. 1254 Davidson, J. L. 5125 Stefanova, Ivanka 6649 Williams, John W. 6649 McLauchlan, Kendra K. 6650 McLauchlan, Kendra K. 6651 Commerford, Julie L. 6652 Morris, Christopher J. 7604 Sugita, Shinya 10319 Parker, Nancy E. 11 Baker, Richard G. 28 Baker, Richard G. 32 Davis, Anthony M. 33 Delcourt, Hazel R. 44 Delcourt, Paul A. 53 Grimm, Eric Christopher 63 Kapp, Rold O. 67 Lichti-Federovich, Sigrid 69 Maher, Louis J., Jr. 74 McAndrews, John H. 82 Mott, Robert J. 86 Peters, M. A. 99 Ritchie, James C. 101 Swain, Patricia C. 101 Van Zant, Kent L. 106 Van Zant, Kent L. 108 Watts, William A. 108 Webb, Thompson III 114 Webb, Thompson III 125 Wright, Herbert E., Jr. 136 Chumbley, Craig A. 146 Hall, Stephen A. 163 MacDonald, Glen 163 Peterson, G. M. 685 Peterson, G. M. 1254 Davidson, J. L. 5125 Stefanova, Ivanka

Table S 2-2 (continued)

Contact ID Contact Name 6649 Williams, John W. 6649 McLauchlan, Kendra K. 6650 McLauchlan, Kendra K. 6651 Commerford, Julie L. 6652 Morris, Christopher J. 7604 Sugita, Shinya 10319 Parker, Nancy E.

Table S 2-3: Citation information for all presettlement pollen datasets for which citation information is available.

Dataset Citation ID Maher, L.J., Jr. 1982. The palynology of Devils Lake, Sauk County, Wisconsin. Pages 119-135 in J.C. Knox, L. Clayton, and D.M. Mickelson, editors. Quaternary History of the Driftless Area. Field 7 Trip Guide Book 5. University of Wisconsin-Extension, Geological and Natural History Survey, Madison, Wisconsin, USA. Notes: Prepared for 29th Annual Meeting Midwest Friends of the Pleistocene, Prairie du Chien, Wisconsin, May 22 and 23, 1982. Davis, A.M. 1977. The prairie-deciduous forest ecotone in the upper Middle West. Annals of the 38 Association of American Geographers 67(2):204-213. Grimm, E.C. 1983. Chronology and dynamics of vegetation change in the prairie-woodland region 54 of southern Minnesota, U.S.A. New Phytologist 93(2):311-350. [DOI: 10.1111/j.1469- 8137.1983.tb03434.x] Albert, L.E., and D.G. Wyckoff, editors. 1981. Ferndale Bog and Natural Lake: five thousand years 67 of environmental change in southeastern Oklahoma. Studies in Oklahoma's Past 7. Oklahoma Archaeological Survey, Norman, Oklahoma, USA. Mott, R.J. 1973. Palynological studies in central Saskatchewan: pollen stratigraphy from lake 74 sediment sequences. Paper 72-49. Geological Survey of Canada, Ottawa, Ontario, Canada. Ritchie, J.C., and K. Hadden. 1975. Pollen stratigraphy of the Holocene sediments from the Grand 77 Rapids area, Manitoba, Canada. Review of Palaeobotany and Palynology 19(3):193-202. [DOI: 10.1016/0034-6667(75)90040-8] Ritchie, J.C. 1964. Contributions to the Holocene paleoecology of west central Canada. 1. The 78 Riding Mountain area. Canadian Journal of Botany 42(2):181-196. Van Zant, K.L. 1979. Late glacial and postglacial pollen and plant macrofossils from Lake West 85 Okoboji, northwestern Iowa. Quaternary Research 12(3):358-380. [DOI: 10.1016/0033- 5894(79)90034-6] Wright, H.E., Jr., T.C. Winter, and H.L. Patten. 1963. Two pollen diagrams from southeastern Minnesota: problems in the regional late-glacial and postglacial vegetation history. Geological 93 Society of America Bulletin 74(11):1371-1396. [DOI: 10.1130/0016- 7606(1963)74[1371:TPDFSM]2.0.CO;2] Wright, H.E., Jr., and W.A. Watts. 1969. Glacial and vegetation history of northeastern Minnesota. 94 Special Publications Series 11. Minnesota Geological Survey, University of Minnesota, Minneapolis, Minnesota, USA. Ritchie, J.C., and S. Lichti-Federovich. 1968. Holocene pollen assemblages from the Tiger Hills, 129 Manitoba. Canadian Journal of Earth Sciences 5:873-880. McAndrews, J.H. 1982. Holocene environment of a fossil bison from Kenora, Ontario. Ontario 140 Archaeology 37:41-51. Brugam, R.B., E.C. Grimm, and N.M. Eyster-Smith. 1988. Holocene environmental changes in Lily 158 Lake, Minnesota inferred from fossil diatom and pollen assemblages. Quaternary Research 30(1):53-66. [DOI: 10.1016/0033-5894(88)90087-7]

Table S 2-3 (continued)

Dataset Citation ID Kim, H. K. 1986. Late-glacial and Holocene environment in central Iowa: a comparative study of 160 pollen data from four sites. Dissertation. University of Iowa, Iowa City, Iowa, USA. Stuiver, M. 1969. Yale Natural Radiocarbon Measurements IX. Radiocarbon 11(2):545-658. [DOI: 168 10.1017/S0033822200011413] Grimm, E.C. 1981. An ecological and paleoecological study of the vegetation in the Big Woods 173 region of Minnesota. Doctoral dissertation. University of Minnesota, Minneapolis, Minnesota, USA. Almquist-Jacobson, H., J.E. Almendinger, and S.E. Hobbie. 1992. Influence of terrestrial vegetation 176 on sediment-forming processes in kettle lakes of west-central Minnesota. Quaternary Research 38(1):103-116. MacDonald, G.M. 1989. Postglacial palaeoecology of the subalpine forest–grassland ecotone of southwestern Alberta: new insights on vegetation and climate change in the Canadian Rocky 196 Mountains and adjacent foothills. Palaeogeography, Palaeoclimatology, Palaeoecology 73(3-4):155- 173. [DOI: 10.1016/0031-0182(89)90001-1] Baker, R.G., C.A. Chumbley, P.M. Witinok, and H.K. Kim. 1990. Holocene vegetation changes in 199 eastern Iowa. Journal of the Iowa Academy of Science 97:167-177. Baker, R.G., L.J. Maher, Jr., C.A. Chumbley, and K.L. Van Zant. 1992. Patterns of Holocene 202 environmental change in the midwestern United States. Quaternary Research 37(3):379-389. [DOI: 10.1016/0033-5894(92)90074-S] Barnosky, C.W., E.C. Grimm, and H.E. Wright, Jr. 1987. Towards a postglacial history of the 224 northern Great Plains: a review of the paleoecologic problems. Annals of the Carnegie Museum 56:259-273. McAndrews, J.H. 1968. Pollen evidence for the protohistoric development of the "Big Woods" in 256 Minnesota (U.S.A.). Review of Palaeobotany and Palynology 7:201-211. McAndrews, J.H. 1966. Postglacial history of prairie, savanna, and forest in northwestern 265 Minnesota. Torrey Botanical Club Memoir 22(2):1-72. Almendinger, J.C. 1992. The late Holocene history of prairie, brush-prairie, and jack pine (Pinus 269 banksia) forest on outwash plains, north-central Minnesota, USA. The Holocene 2(1):37-50. [DOI: 10.1177/095968369200200105] McAndrews, J.H. 1969. Paleobotany of a wild rice lake in Minnesota. Canadian Journal of Botany 270 47:1671-1679. McAndrews, J.H. 1988. Human disturbance of North American forests and grasslands: the fossil 271 pollen record. Pages 673-697 in B. Huntley and T. Webb III, editors. Vegetation History. Kluwer Academic Publishers, Dordrecht, Netherlands. Stuiver, M. 1975. Climate versus changes in 13C content of the organic component of lake 332 sediments during the late Quaternary. Quaternary Research 5(2):251-262. [DOI: 10.1016/0033- 5894(75)90027-7] Gajewski, K., editor. 1985. Late-Holocene pollen data from lakes with varved sediments in 394 northeastern and northcentral United States. IES Report 124, Center for Climatic Research, University of Wisconsin, Madison, Wisconsin, USA. Swain, P.C. 1979. The development of some bogs in eastern Minnesota. Doctoral dissertation. 402 University of Minnesota, Minneapolis, Minnesota, USA. Murchie, S.L. 1984. 210Pb dating and recent geologic history of two bays of Lake Minnetonka, 437 Minnesota. Thesis. University of Minnesota, Minneapolis, Minnesota, USA. Lowdon, J.A., and W. Blake, Jr. 1979. Geological Survey of Canada radiocarbon dates XIX. Paper 460 79-7, Geological Survey of Canada. Haworth, E.Y. 1972. Diatom succession in a core from Pickerel Lake, northeastern South Dakota. 468 Geological Society of America Bulletin 83(1):157-172. [DOI: 10.1130/0016- 7606(1972)83[157:DSIACF]2.0.CO;2] Jacobson, G.L., Jr. 1975. A palynological study of the history and ecology of white pine in 473 Minnesota. Doctoral dissertation. University of Minnesota, Minneapolis, Minnesota, USA.

Table S 2-3 (continued)

Dataset Citation ID Alwin, B.C. 1982. Vegetation history of the Sugar Hills area, Itasca Co., Minnesota. Master's thesis. 475 University of Minnesota, Minneapolis, Minnesota, USA. McAndrews, J.H. 1967. Pollen analysis and vegetation history of the Itasca region, Minnesota. 505 Pages 219-236 in E.J. Cushing and H.E. Wright, Jr., editors. Quaternary Paleoecology. Yale University Press, New Haven, Connecticut, USA. Allen, K.E. 1992. Geochemical stratigraphy of Medicine Lake, South Dakota: Paleoclimatic 514 implications of the freshwater/ saline transition. Thesis. University of Minnesota, Minneapolis, Minnesota, USA. Laird, K.R. 1996. A high-resolution paleoclimatic record of a closed-basin lake in the northern 597 Great Plains. Doctoral dissertation. University of Minnesota, Minneapolis, Minnesota, USA. Ritchie, J.C. 1976. The late-Quaternary vegetation history of the western interior of Canada. 630 Canadian Journal of Botany 54(15):1793-1818. [DOI: 10.1139/b76-194] Vance, R.E., W.M. Last, and A.J. Smith. 1997. Hydrologic and climatic implications of a 688 multidisciplinary study of late Holocene sediment from Kenosee Lake, southeastern Saskatchewan, Canada. Journal of Paleolimnology 18(4):365-393. [DOI: 10.1023/A:1007948909342] Lichti-Federovich, S. 1970. The pollen stratigraphy of a dated section of late Pleistocene lake 970 sediment from central Alberta. Canadian Journal of Earth Sciences 7:938-945. Patterson, W.A., III. 1978. The effects of past and current land disturbances on Squaw Lake, 7223 Minnesota and its watershed. Doctoral dissertation. University of Minnesota, St. Paul, Minnesota, USA. Lynch, E.A., S.C. Hotchkiss, and R. Calcote. 2011. Charcoal signatures defined by multivariate 7275 analysis of charcoal records from 10 lakes in northwest Wisconsin (USA). Quaternary Research 75(1):125-137. [DOI: 10.1016/j.yqres.2010.08.007] Hotchkiss, S.C., R. Calcote, and E.A. Lynch. 2007. Response of vegetation and fire to Little Ice Age 7280 climate change: regional continuity and landscape heterogeneity. Landscape Ecology 22:25-41. [DOI: 10.1007/s10980-007-9133-3] Tweiten, M.A., S.C. Hotchkiss, R.K. Booth, R. Calcote, and E.A. Lynch. 2009. The response of a 7282 jack pine forest to late-Holocene climate variability in northwestern Wisconsin. The Holocene 19(7):1049-1061. Camill, P., C.E. Umbanhowar, Jr., R. Teed, C.E. Geiss, J. Aldinger, L. Dvorak, J. Kenning, J. Limmer, and K. Walkup. 2003. Late-glacial and Holocene climatic effects on fire and vegetation 7293 dynamics at the prairie-forest ecotone in south- central Minnesota. Journal of Ecology 91(5):822- 836. 10.1046/j.1365-2745.2003.00812.x. Geiss, C.E., C.E. Umbanhowar, P. Camill, and S.K. Banerjee. 2003. Sediment magnetic properties 7294 reveal Holocene climate change along the Minnesota prairie-forest ecotone. Journal of Paleolimnology 30(2):151-166. [DOI: 10.1023/A:1025574100319] Jaumann, P.J. 1995. Century to millennium-scale late Quaternary Natural climate variability in the 7319 Midwestern United States. Doctoral dissertation. University of Iowa, Iowa City, Iowa, USA. Notes: Thesis supervisor: Richard G. Baker Ollendorf, A.L. 1993. Changing landscapes in the American Bottom (USA): An interdisciplinary 7374 investigation with an emphasis on the late-prehistoric and early-historic periods. Doctoral dissertation. University of Minnesota, Minneapolis, Minnesota, USA. Umbanhowar, C.E., Jr. 2004. Interaction of fire, climate and vegetation change at a large landscape 7389 scale in the Big Woods of Minnesota, USA. The Holocene 14(5):661-676. [DOI: 10.1191/0959683604hl745rp] Munoz, S.E., S. Schroeder, D.A. Fike, and J.W. Williams. 2014. A record of sustained prehistoric 7413 and historic land use from the Cahokia region, Illinois, USA. Geology 42(6):499-502. http://geology.gsapubs.org/content/42/6/499.short. [DOI: 10.1130/G35541.1] St. Jacques, J.-M., B.F. Cumming, and J.P. Smol. 2008. A 900-year pollen-inferred temperature and 7571 effective moisture record from varved Lake Mi, west-central Minnesota, USA. Quaternary Science Reviews 27(7-8):781-796. [DOI: 10.1016/j.quascirev.2008.01.005]

Table S 2-3 (continued)

Dataset Citation ID Brooks, G.R., and A. Grenier. 2001. Late Holocene pollen stratigraphy of Lake Louise, Manitoba. 7663 Current Research 2001-B1, Geological Survey of Canada, Ottawa, Ontario, Canada. Davis, M., C. Douglas, R. Calcote, K.L. Cole, M.G. Winkler, and R. Flakne. 2000. Holocene 7674 climate in the western Great Lakes national parks and lakeshores: implications for future climate change. Conservation Biology 14(4):968-983. [DOI: 10.1046/j.1523-1739.2000.99219.x] Jacobson, H.A., and D.R. Engstrom. 1989. Resolving the chronology of recent lake sediments: an 7705 example from Devils Lake, North Dakota. Journal of Paleolimnology 2(2):81-97. [DOI: 10.1007/BF00177042] Teed, R.E., C.E. Umbanhowar, and P. Camill. 2009. Multiproxy lake sediment records at the 7707 northern and southern boundaries of the Aspen Parkland region of Manitoba, Canada. The Holocene 19(6):937-948. [DOI: 10.1177/0959683609336569] Locke, S.M. 1995. A paleohydrologic model applied to the Holocene sediment stratigraphy of two 7766 lakes in north-central Minnesota. Doctoral dissertation. University of Minnesota, Minneapolis, Minnesota, USA. Young, M.A. 1979. A high resolution paleomagnetic study of recent sediments from Long Lake, 7774 Hennepin County, Minnesota. Master's thesis. University of Minnesota, Minneapolis, Minnesota, USA. Spigel, K.M. 2006. Erosion and sedimentation history of Emrick Lake, south-central Wisconsin, in 7776 response to Holocene environmental change. Doctoral dissertation. University of Wisconsin, Madison, Wisconsin, USA. Hu, F.S., D. Slawinski, H.E. Wright, Jr., E. Ito, R.G. Johnson, K.R. Kelts, R.F. McEwan, and A. 7835 Boedigheimer. 1999. Abrupt changes in North American climate during early Holocene times. Nature 400:437-440. [DOI: 10.1038/2272] Nichols, H. 1969. The late Quaternary history of vegetation and climate at Porcupine Mountain and 7907 Clearwater Bog, Manitoba. Arctic and Alpine Research 1(3):155-167. Commerford, J.L., B. Leys, J.R. Mueller, and K.K. McLauchlan. 2016. Great Plains vegetation 9415 dynamics in response to fire and climatic fluctuations during the Holocene at Fox Lake, Minnesota (USA). The Holocene 26(2):302-313. Brown, K.J., J.S. Clark, E.C. Grimm, J.J. Donovan, P.G. Mueller, B.C.S. Hansen, and I. Stefanova. 2005. Fire cycles in North American interior grasslands and their relation to prairie drought. 9523 Proceedings of the National Academy of Sciences of the United States of America 102(25):8865- 8870. [DOI: 10.1073/ps.0503621102] Campbell, C., I.D. Campbell, C.B. Blyth, and J.H. McAndrews. 1994. Bison extirpation may have 9587 caused aspen expansion in western Canada. Ecography 17(4):360-362. Munoz, S.E., K.E. Gruley, A. Massie, D.A. Fike, S. Schroeder, and J.W. Williams. 2015. Cahokia's 9856 emergence and decline coincided with shifts of flood frequency on the Mississippi River. Proceedings of the National Academy of Sciences 112(20):6319-6324. Hart, C.L. 2009. High-resolution pollen analysis of two lakes at the boreal forest-aspen parkland 10603 ecotone in central Saskatchewan, Canada. Master's thesis. University of Regi, Regi, Saskatchewan, Canada.

Table S 2-4: Dataset ID and contact name for all surface pollen datasets for which full citation information is not available.

Contact ID Contact Name 4 Albert, Lois E. 6 Almendinger, John C. 11 Baker, Richard G. 28 Davis, Anthony M. 29 Davis, Margaret B. 34 Engstrom, Daniel R. 35 Eyster-Smith, Nancy M. 40 Gajewski, Konrad J. 44 Grimm, Eric Christopher 45 Hadden, Kathleen A. 46 Hansen, Barbara C. S. 50 Jacobson, George L., Jr. 57 Kim, H. K. 63 Lichti-Federovich, Sigrid 67 Maher, Louis J., Jr. 69 McAndrews, John H. 76 Nichols, Harvey 81 Patterson, William Albert, III 86 Ritchie, James C. 99 Swain, Patricia C. 101 Van Zant, Kent L. 106 Watts, William A. 112 Winkler, Marjorie Green 113 Winter, Thomas C. 114 Wright, Herbert E., Jr. 115 Wyckoff, Don G. 123 Brugam, Richard Blair 125 Chumbley, Craig A. 126 Clark, James S. 127 Cole, Kenneth L. 138 Hobbie, Sarah E. 142 Jaumann, Peter Joseph 146 MacDonald, Glen 161 Whitlock, Cathy 201 Patten, Harvey L. 206 Stuiver, Minze 229 Almendinger, James E. 230 Almquist, Heather R. 264 Vance, Robert E. 271 Witinok, P. M. 273 Campbell, Ian D. 305 Hu, Feng Sheng 314 Smol, John P. 378 Blake, Weston J., Jr. 382 Lowdon, J. A. 469 Murchie, S. L. 490 Haworth, Elizabeth Y. 493 Alwin, Baird Christopher 532 Allen, K. E.

Table S 2-4 (continued)

Contact ID Contact Name 605 Mueller, Pietra G. 606 Laird, Kathleen R. 608 Cumming, Brian F. 683 Last, William M. 684 Smith, Alison Jean 746 Lynch, Elizabeth A. 941 Brown, Kendrick J. 1254 Stefanova, Ivanka 1484 Ito, Emi 1495 Kelts, Kerry R. 4807 Campbell, Celi 5125 Williams, John W. 6380 Calcote, Randy 6381 Hotchkiss, Sara C. 6389 Tweiten, Michael A. 6390 Booth, Robert K. 6403 Umbanhowar, Charles E., Jr. 6404 Camill, Philip 6405 Teed, Rebecca E. 6406 Geiss, Christoph E. 6407 Banerjee, Subir K. 6408 Aldinger, Jessica 6409 Dvorak, Leah 6410 Kenning, Jon 6411 Limmer, Jacob 6412 Walkup, Kristi 6438 Donovan, Joseph 6530 Ollendorf, Amy L. 6583 Munoz, Sam 6584 Schroeder, Sissel 6585 Fike, David 6649 McLauchlan, Kendra K. 6650 Commerford, Julie L. 6782 St. Jacques, Jeannine-Marie 6868 Brooks, Greg R. 6869 Grenier, A. 6882 Douglas, Christine 6883 Flakne, Robyn A. 6934 Mott, Robert J. 6973 Locke, Sharon M. 6983 Young, Michael Allen 6984 Spigel, Kevin M. 7105 Boedigheimer, A. 7106 Johnson, R.G. 7107 Slawinski, D. 7108 McEwan, R.F. 8681 Leys, Berangere 8683 Mueller, Joshua Robert 8931 Blyth, C. B. 9222 Massie, Ashtin

Table S 2-4 (continued)

Contact ID Contact Name 9223 Gruley, Kristine E. 9888 Hart, Catherine Leigh 12553 Jacobson, Heather

Table S 2-5: Number of pollen datasets per ecoregion in the surface set used in ecoregion-based analyses. Ecoregion Surface Set Boreal plain 56 Central USA plains 15 MS alluvial and SE USA coastal plains 10 Mixed wood plains 92 Mixed wood shield 26 Ozark/Ouachita-Appalachian forests 11 South central semiarid prairies 55 Southeastern USA plains 13 Temperate prairies 85 West-central semiarid prairies 132 Total 497

Figure S 2-1: Bootstrap results from 1,000 estimates of the R-squared (t) associated with a B- spline regression with three knots. The Ambrosia to Artemisia ratio from the "Both Present" surface dataset is regressed against MAP. The Ambrosia to Artemisia ratio was log transformed prior to model fitting. Original R2 = 0.777, bias = 0.002, std. error = 0.021

Figure S 2-2: Model fit check for the B-spline regression with three knots developed for the “Both Present” surface dataset. The Ambrosia to Artemisia ratio from the "Both Present" surface dataset is regressed against MAP. The Ambrosia to Artemisia ratio was log transformed prior to model fitting.

Table S 2-6: Pairwise ANOVA results comparing the Ambrosia to Artemisia ratio between ecoregions. P-values are adjusted for multiple comparisons using Tukey's HSD. Pairs which are significantly different at the α = 0.05 level after adjustment are given in bold. Ecoregion Pair Difference Lower Upper Adjusted p-value Central USA Plains-Boreal Plain 11.67 8.18 15.15 < 1x10-7 Mississippi Alluvial and Southeast USA Coastal Plains-Boreal Plain 14.45 10.33 18.56 < 1x10-7 Mixed Wood Plains-Boreal Plain 12.47 10.44 14.50 < 1x10-7 Mixed Wood Shield-Boreal Plain 7.63 4.79 10.48 < 1x10-7 Ozark/Ouachita-Appalachian Forests-Boreal Plain 13.16 9.21 17.11 < 1x10-7 South Central Semiarid Prairies-Boreal Plain 8.82 6.56 11.07 < 1x10-7 Southeastern USA Plains-Boreal Plain 13.16 9.47 16.85 < 1x10-7 Temperate Prairies-Boreal Plain 5.73 3.67 7.79 < 1x10-7 West-Central Semiarid Prairies-Boreal Plain 3.60 1.69 5.51 1.78x10-7 Mississippi Alluvial and Southeast USA Coastal Plains-Central USA Plains 2.78 -2.11 7.67 0.73

54 Mixed Wood Plains-Central USA Plains 0.80 -2.53 4.14 1

Mixed Wood Shield-Central USA Plains -4.03 -7.92 -0.15 0.03 Ozark/Ouachita-Appalachian Forests-Central USA Plains 1.49 -3.26 6.25 0.99 South Central Semiarid Prairies-Central USA Plains -2.85 -6.33 0.63 0.22 Southeastern USA Plains-Central USA Plains 1.49 -3.05 6.03 0.99 Temperate Prairies-Central USA Plains -5.94 -9.29 -2.58 1.41x10-6 West-Central Semiarid Prairies-Central USA Plains -8.06 -11.33 -4.80 < 1x10-7 Mixed Wood Plains-Mississippi Alluvial and Southeast USA Coastal Plains -1.98 -5.97 2.01 0.86 Mixed Wood Shield-Mississippi Alluvial and Southeast USA Coastal Plains -6.81 -11.27 -2.35 6.99x10-5 Ozark/Ouachita-Appalachian Forests-Mississippi Alluvial and Southeast USA Coastal Plains -1.29 -6.52 3.95 1 South Central Semiarid Prairies-Mississippi Alluvial and Southeast USA Coastal Plains -5.63 -9.74 -1.52 0.0007 Southeastern USA Plains-Mississippi Alluvial and Southeast USA Coastal Plains -1.29 -6.33 3.75 1 Temperate Prairies-Mississippi Alluvial and Southeast USA Coastal Plains -8.72 -12.72 -4.71 < 1x10-7 West-Central Semiarid Prairies-Mississippi Alluvial and Southeast USA Coastal Plains -10.84 -14.77 -6.91 < 1x10-7 Mixed Wood Shield-Mixed Wood Plains -4.84 -7.50 -2.17 6.10x10-7

Table S 2-6 (continued)

Ecoregion Pair Difference Lower Upper Adjusted p-value Ozark/Ouachita-Appalachian Forests-Mixed Wood Plains 0.69 -3.13 4.51 1 South Central Semiarid Prairies-Mixed Wood Plains -3.65 -5.67 -1.63 7.24x10-7 Southeastern USA Plains-Mixed Wood Plains 0.69 -2.86 4.24 1 Temperate Prairies-Mixed Wood Plains -6.74 -8.54 -4.94 < 1x10-7 West-Central Semiarid Prairies-Mixed Wood Plains -8.87 -10.49 -7.24 < 1x10-7 Ozark/Ouachita-Appalachian Forests-Mixed Wood Shield 5.53 1.22 9.84 0.002 South Central Semiarid Prairies-Mixed Wood Shield 1.19 -1.65 4.02 0.95 Southeastern USA Plains-Mixed Wood Shield 5.52 1.45 9.60 0.0008 Temperate Prairies-Mixed Wood Shield -1.90 -4.59 0.78 0.42 West-Central Semiarid Prairies-Mixed Wood Shield -4.03 -6.60 -1.46 3.83x10-5 South Central Semiarid Prairies-Ozark/Ouachita-Appalachian Forests -4.34 -8.29 -0.40 0.02 Southeastern USA Plains-Ozark/Ouachita-Appalachian Forests 0.00 -4.91 4.91 1

55 -7 Temperate Prairies-Ozark/Ouachita-Appalachian Forests -7.43 -11.27 -3.59 < 1x10 West-Central Semiarid Prairies-Ozark/Ouachita-Appalachian Forests -9.56 -13.32 -5.80 < 1x10-7 Southeastern USA Plains-South Central Semiarid Prairies 4.34 0.66 8.02 0.008 Temperate Prairies-South Central Semiarid Prairies -3.09 -5.14 -1.04 9.81x10-5 West-Central Semiarid Prairies-South Central Semiarid Prairies -5.22 -7.11 -3.32 < 1x10-7 Temperate Prairies-Southeastern USA Plains -7.43 -11.00 -3.86 < 1x10-7 West-Central Semiarid Prairies-Southeastern USA Plains -9.55 -13.04 -6.07 < 1x10-7 West-Central Semiarid Prairies-Temperate Prairies -2.13 -3.79 -0.46 0.002

CHAPTER 3. GEOPOLLEN: AN INTERACTIVE, USER-FRIENDLY SHINY DASHBOARD APPLICATION FOR VISUALIZATION AND ANALYSIS OF POLLEN DATA

Hannah M. Carroll1, Alan D. Wanamaker2, and Lynn G. Clark1

1Department of Ecology Evolution, and Organismal Biology, Iowa State University, Ames,

Iowa, United States of America; 2Department of Geological and Atmospheric Sciences, Iowa

State University, Ames, Iowa, United States of America

Modified from a manuscript to be submitted to PLOS One

Abstract

Pollen data are routinely used in paleoenvironmental studies to understand past climate and vegetation. A large and active community contributes to and maintains the Neotoma

Paleoecology Database, a massive repository for a wide variety of paleo proxy data. One existing limitation in working with pollen data is the need to write code in order to execute several of the routine analyses in paleoecological work. We developed GeoPollen, a Shiny Dashboard application, to be a streamlined, user-friendly GUI-based tool for performing these basic analyses. Users are able to quickly search, visualize, analyze, and download more than 3,000 publicly available pollen datasets from the Neotoma Database spanning the last 22,000 years across the United States and Canada. We developed GeoPollen in order to improve the openness and accessibility of the Neotoma Database, which is central to the project’s mission.

Introduction

Pollen data are widely employed in paleoecological and paleoclimatological studies to reconstruct past vegetation and climate. A large, global, and highly active community of researchers contribute to an ever-growing body of knowledge. Much of this knowledge has been made readily accessible by the advent of the Neotoma Paleoecology Database [1] and its

57 constituent databases, including the North American Pollen Database [2]. Neotoma holds well over 3,000,000 pollen records [1] representing over 8,000 individual datasets.

Neotoma provides two GUI-based tools for directly interacting with pollen data:

Neotoma Explorer and Tilia. Neotoma Explorer allows users to directly search for and download pollen data, view raw data and metadata, and create visually appealing stratigraphy plots. Tilia allows users to download data, perform cluster analyses, develop age models, create plots, and prepare their own data for upload [3]. Most tools currently available to researchers for processing and visualizing pollen data beyond these tasks rely on the user’s ability to write computer code, typically in R. The most commonly used of these are R packages ‘neotoma’ [4] and ‘analogue’

[5,6]. Package ‘neotoma’ provides a link to the Neotoma Database via API calls, wherein users can search for, download, standardize, and plot data. Package ‘analogue’, which is separate from the Neotoma suite of software, provides tools for doing many commonly employed analyses in paleoclimatology and paleoecology. Although these packages are powerful, fast, and flexible, the necessity of writing code imposes a point-of-entry barrier for many researchers. To address this barrier, we have developed a free, publicly available, open-source application that allows users to execute some of the most common analyses in paleoecology without the need to write code.

GeoPollen is a user-friendly GUI interface developed as an R Shiny Dashboard. The dashboard allows users to perform several of the most frequently employed analyses via a simple interface which requires no expertise in coding, and no software to be downloaded. GeoPollen does not replicate any functions currently available in Neotoma’s own suite of tools aside from the ability to search for and download data. GeoPollen is not associated with the Neotoma

Paleoecology Database or its constituent databases, and is meant simply as an external tool for users to perform basic analyses on publicly available pollen data.

If desired, users can clone GeoPollen’s GitHub repository and run a local instance of the

GeoPollen application in RStudio, or reuse the pre-built code in their own analyses. The data available in GeoPollen span the continental US and Canada from 20,000 calendar years before present to the present and are updated monthly via an application programming interface (API) call to the Neotoma Paleoecology Database in R package ‘neotoma’ [4].

Materials and methods

GeoPollen is a Shiny Dashboard application developed using package ‘shinydashboard’ version 0.7.1 [7] in R version 3.6.2 [8]. It is hosted on a webserver located at https://geopollen.eeob.iastate.edu/ and can be executed in the user’s browser without any software downloads required. Shiny Dashboards automatically scale to the user’s web browser; however, users may zoom in or out within their browser to rescale the interface according to their viewing preferences. If desired, the source code and data can be cloned from GitHub

(https://github.com/hannahcarroll/GeoPollen) and run on the user’s local machine in RStudio [9].

Code may be repurposed or modified by users familiar with the R language to fit their individual requirements. GeoPollen leverages fossil pollen datasets available through the Neotoma

Paleoecology Database [1]. To maintain speed and reactivity in the application, available datasets are currently limited to just fossil pollen (not surface samples) in the US and Canada spanning the Last Glacial Maximum (approximately 22 kya) to the present. Future releases will expand data availability as processing capabilities improve.

Data pre-processing

Raw data are pre-processed at application launch in a global.R script running silently behind GeoPollen. This is enables faster reaction times in individual user sessions. Pollen are first standardized to the Whitmore Full list [10] using package ‘neotoma’ [4]. Non-pollen palynomorphs and spikes (i.e., microspheres, Lycopodium tablets, etc.) are dropped. Raw pollen

59 counts are then converted to relative proportions. If a user wishes to download the raw count data without non-pollen palynomorphs (e.g., fungal spores) and spikes (e.g., Lycopodium tablets, microspheres, etc.) dropped, and without the Whitmore Full standardization, they may do so through download buttons located in either the Main or Distribution Mapping tabs or directly from the Neotoma Paleoecology Database. If a user wishes to keep these standardizations, they may copy and paste datasets from within the app itself or re-use the underlying source code directly.

Raw radiocarbon ages submitted by dataset authors are calibrated to IntCal09 [11] at app deployment and stored in a new column named “calendar.ybp.” Where authors have submitted calibrated radiocarbon dates, no changes are made and the information is copied to the

“calendar.ybp” column. The original columns defining age, minimum age, maximum age, and date type as submitted by dataset authors are maintained.

Discussion of the GeoPollen interface

The GeoPollen user interface is divided into sidebar tabs to streamline the user’s workflow (Figure 3-1). Reactive inputs and outputs managed by package ‘shiny’ [12] allow for on-the-fly data manipulation. Breaking tasks across tabs and subtabs improves application

response times for large datasets by calling only the necessary functions

for the user’s current operation and minimizing graphics processing

demands.

Figure 3-1: GeoPollen’s Sidebar. GeoPollen’s user interface is divided into tabs, organized into a sidebar. Each item contains various functions meant to streamline the user’s workflow.

The Main tab

The Main tab reactively displays the locations of available datasets (Figure 3-2). Circle markers represent individual sites, each of which reports one or more rows of pollen count data, or datasets. Required input from the user is an age range in calendar years before present for which to display available data. As the age slider is manipulated, the available datasets are

Figure 3-2: GeoPollen’s Main Tab. The Main tab reactively displays the location of available data in the user’s selected age range. Optional search parameters include geographic region, state/province, or site name. The download button sends a call to the Neotoma database via package ‘neotoma’. reactively subset and dataset locations are displayed on a Leaflet [13] map via a Leaflet proxy call. The user has the option to select one or more datasets by site name, state/province, or geographic region, but this input is not required. Functions within GeoPollen reactively subset data as the user moves or zooms the map. Action buttons allow the user to reset all selections and return the map to default, or to download a spreadsheet in .csv format of all raw data from all sites currently in view on the map. This is achieved via an API call to the Neotoma Paleoecology

Database implemented in package ‘neotoma’ [4].

A histogram of visible dataset ages updates as the user makes selections, and a table provides summaries of the ages of currently visible datasets, the number of individual sites displayed, and the total number of rows of data (Figure 3-3). The data summaries underlying both the histogram and table are produced using grouping and summarizing functions available

61 in package ‘dplyr’ [14]. Action buttons allow all selections to be cleared and for the map to be reset to its default state, or for all currently visible data to be downloaded in tabular (.csv) format.

Figure 3-3: Visible Dataset Summaries. A histogram and table populated by the datasets currently visible on the map

provide quick summaries on which users can base processing decisions.

Modern Analog Technique calculations

One of the most commonly performed analyses in paleoecology is the Modern Analog

Technique (MAT) [6]. GeoPollen employs R package ‘analogue’ [5] to allow the user to perform basic analog calculations on the fly. The user selects a map marker, representing a site, i.e., the

physical location from which a

record was derived, to bring up an

additional dialogue in the Analog

Controls box in the Main tab

(Figure 3-4). The user then selects

an age range in this dataset to serve

as the reference, or training set. All Figure 3-4: The Analog Controls Box. Clicking a site on the map brings up additional options in the Analog other datasets in the user’s selected Controls box. The scatterplot shows an age-depth plot for the selected site within the selected age range. The user spatiotemporal range are used at may select some or all of the rows of data to serve as the training set for analog calculations. the test set. Pressing the Find

Analogs button triggers the analog function from package ‘analogue’ [5] to calculate the Squared

Chord Distance (SCD) between the training set and all rows of the test set. Diagnostics are returned to the Analog Communities tab and its sub-tabs. Although multiple distance metrics are available through package ‘analogue’ we have elected to implement SCD only in the first release of GeoPollen. SCD is, at present, the preferred distance metric in pollen analyses and the one most likely to be familiar to users. Future releases of GeoPollen will expand this capability and automate the calculation of appropriate cutoffs for analog vs. no-analog situations.

Analog Communities tab

The training set’s metadata is returned to the Minimum Dissimilarity sub-tab, accompanied by output from the analog function of package ‘analogue’. Default output includes the call sent to the analog function, the dissimilarity metric used (Squared Chord Distance), training set dissimilarities, and the dissimilarity between the training set and each item in the test set (Figure 3-5).

Figure 3-5: Minimum Dissimilarity Output. Training set metadata and default output from the analog function are returned to the Minimum Dissimilarity tab. This allows the user to confirm their selections and view basic information about their chosen data.

The Training Set Diagnostics subtab provides additional information about the user’s training set (Figure 3-6). A graph of expected versus observed distributions is provided by package ‘analogue’. A heatmap displays the SQD between all members of the training set. Both plots are downloadable. Heatmaps are produced in package ‘plotly’ [15] using colorblind- friendly palettes from package ‘viridis’ [16].

Figure 3-6: Training Set Diagnostics. The Training Set Diagnostics subtab displays the distribution of the expected versus observed training set dissimilarities, and a heatmap of training set dissimilarities produced using package ‘plotly’ [15]. The heatmap is fully interactive and users may download either plot in .png format. Datasets are identified by a concatenation of the site name (Roe Lake), dataset ID (multiple), depths, and ages in Cal BP.

The Analogs subtab provides both a fully interactive heatmap and a searchable table of the SCD between each member of the test set and training set (Figure 3-7). Users are able to download the heatmap in .png format and copy a matrix of pairwise SCD directly from the searchable table.

Figure 3-7: Analogs Subtab Output. The Analogs subtab displays a heatmap produced in ‘plotly’ [15] populated with the Squared Chord Distance between the training set (vertical axis) and test set (horizontal axis). The heatmap is fully interactive and users may download the plot in .png format.

Distribution Mapping tab

The ability to quickly visualize the spatiotemporal distribution of vegetation using a large number of datasets is fundamental to understanding past plant communities and climate.

GeoPollen’s Distribution tab responds to user input to display all reported instances of one or more taxa over the spatiotemporal range selected by the user. Functionality of this tab is similar to that currently offered by the GUI-based Neotoma Explorer, but provides additional flexibility in search options. The key difference is that GeoPollen is focused on end-user analyses, rather than replicating the database functionality of Neotoma. Download requests are routed to the

Neotoma database via the API calls implemented in package ‘neotoma’. However, a copy of the

North American portion of the database is packaged within GeoPollen’s Docker container to improve response times.

Taxa are color coded and locations at which they are reported are represented by semi- transparent circles (Figure 3-8). Darker-colored circles indicate more datasets (i.e., more rows of data) containing a particular taxon at a given site. Where two or more taxa overlap, the color of the circles blends between those being represented. A reactive table tracks the spatial range of all taxa currently included in the search over the given time period, the total number of datasets reporting a given taxon as nonzero, and the total number of sites reporting a given taxon. A second reactive table tracks the same information just for the current map view. Download buttons enable to download data for either the full spatial distribution of their selected taxa or just those currently in view on the map.

Figure 3-8: Example Search in the Distribution Mapping Tab. Example search for Cyperaceae (truncated to Cyperace) and Caprifoliaceae (truncated to Caprifolia). Where both taxa are reported at the same site, the marker color blends between the two. Reactive tables display the spatial range and number of occurrences of all taxa included in the search.

Users can view the underlying data in long-format searchable tables in the Raw Data tab.

They can download the underlying data in wide format from the panel of controls, with the options being to download only datasets currently in view on the map, or all datasets. Users may also copy and paste data directly from the Raw Data tabs, but this is only practical for small numbers of datasets.

Raw Data tab

The Raw Data tab is divided into two sub-tabs; one for data visible in the Main tab, and one for data visible in the Distributions tab. Both tabs contain searchable, sortable tables reactively populated with the datasets currently in view on the corresponding map. Data from the

Main tab are displayed in wide format, i.e., with datasets as rows and all pertinent information as columns, because the purpose is to provide site-level information. Data from the Distributions tab are displayed in long format, i.e., with taxon records as rows and dataset information as columns, because the purpose is to provide taxon-specific information.

Other tabs

The Citations tab contains general information for appropriate attribution of the Neotoma

Paleoecology Database and GeoPollen, as well as citation information for packages used by

GeoPollen to perform key functions. Users are advised to read and abide by the Neotoma

Paleoecology Database’s Data Use and Embargo Policy

(https://www.neotomadb.org/data/category/use). The Contact tab provides contact information for the developer and maintainer of GeoPollen for support or to submit a bug report. Users are requested to submit bug reports through GitHub

(https://github.com/hannahcarroll/GeoPollen/issues). The How to Use GeoPollen tab provides a brief user guide with examples of how to perform searches and complete various tasks. It also provides links to documentation for R packages critical to GeoPollen’s functions.

Conclusions

GeoPollen provides a fast, interactive, GUI-based platform for performing basic analyses on pollen data publicly available through the Neotoma Paleoecology Database. Although GUI- based programs, such as PAST [17] and Tilia [18], are available for analysis of pollen data, none are equipped to import, clean, pre-process, and analyze large numbers of datasets directly from

Neotoma. Streamlined coding allows for calculations to be performed in milliseconds on most datasets. Publication-quality plots may be generated from directly within the application, along with several key diagnostics. GeoPollen allows users with no coding experience to leverage two of the most popular and powerful R packages available for paleoecological analysis. New data uploaded to Neotoma will be incorporated monthly with the ongoing assistance of the Iowa State

University Research IT Department. General updates to the GeoPollen application are expected approximately quarterly and support will be ongoing. A test branch in GeoPollen’s GitHub repository will allow for any changes introduced by updates to R packages to be thoroughly bug tested before release. This tool is meant to forward the Neotoma community’s mission of making paleo data open and accessible to the broadest possible audience.

Acknowledgements

The authors wish to thank Levi Baber of the Iowa State University Research IT

Department for building and maintaining GeoPollen’s Docker container and providing ongoing

IT support, and the Department of Ecology, Evolution, and Organismal Biology at Iowa State

University for providing web hosting for the GeoPollen application. Data were obtained from the

Neotoma Paleoecology Database (https://www.neotomadb.org/) and the North American Pollen

Database. The work of data contributors, data stewards, and the Neotoma community is gratefully acknowledged.

Author contributions

HMC conceived of the idea for the project, wrote the R code for the application and underlying analyses, maintains the underlying data, and wrote the manuscript. LGC and ADW supervised the project, provided feedback, and provided input on the manuscript. All authors read the manuscript and contributed to its development.

Data accessibility

All pollen data available through GeoPollen are housed in the Neotoma Paleoecology

Database (https://www.neotomadb.org/). All GeoPollen source code is available via a public

GitHub repository at https://github.com/hannahcarroll/GeoPollen.

References

1. Williams JW, Grimm EC, Blois JL, Charles DF, Davis EB, Goring SJ, et al. The Neotoma Paleoecology Database, a multiproxy, international, community-curated data resource. Quat Res. 2018;89: 156–177. doi:10.1017/qua.2017.105

2. Grimm EC, Blois J, Giesecke T, Graham R, Smith A, Williams J. Constituent databases and data stewards in the Neotoma Paleoecology Database: History, growth, and new directions. Past Glob Change Mag. 2018;26: 64–65. doi:10.22498/pages.26.2.64

3. Grimm EC. Tilia graph v. 2.0. 2. Ill State Mus Res Collect Cent. 2004.

4. Goring S, Dawson A, Simpson GL, Ram K, Graham RW, Grimm EC, et al. neotoma: A Programmatic Interface to the Neotoma Paleoecological Database. Open Quat. 2015;1. doi:10.5334/oq.ab

5. Simpson, G.L., Okasanen, J. analogue: Analogue matching and Modern Analogue Technique transfer function models. 2019. Available: https://cran.r- project.org/package=analogue

6. Simpson GL. Analogue Methods in Palaeoecology: Using the analogue Package. J Stat Softw. 2007;22. doi:10.18637/jss.v022.i02

7. Chang W, Borges Ribeiro B. shinydashboard: Create Dashboards with “Shiny”. 2018. Available: https://CRAN.R-project.org/package=shinydashboard

8. R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2019. Available: https://www.R-project.org/

9. RStudio Team. RStudio: Integrated Development for R. Boston, MA: RStudio, Inc.; 2019. Available: http://www.rstudio.com/

10. Whitmore J, Gajewski K, Sawada M, Williams JW, Shuman B, Bartlein PJ, et al. Modern pollen data from North America and Greenland for multi-scale paleoenvironmental applications. Quat Sci Rev. 2005;24: 1828–1848. doi:10.1016/j.quascirev.2005.03.005

11. Reimer PJ, Baillie MGL, Bard E, Bayliss A, Beck JW, Blackwell PG, et al. IntCal09 and Marine09 radiocarbon age calibration curves, 0–50,000 year Cal BP. Radiocarbon. 2009;51: 1111–1150.

12. Chang W, Cheng J, Allaire J, Xie Y, McPheron J. shiny: Web Application Framework for R. 2019. Available: https://CRAN.R-project.org/package=shiny

13. Cheng J, Karambelkar B, Xie Y. leaflet: Create Interactive Web Maps with the JavaScript “Leaflet” Library. 2019. Available: https://CRAN.R-project.org/package=leaflet

14. Wickham H, François R, Henry L, Müller K. dplyr: A Grammar of Data Manipulation. 2019. Available: https://CRAN.R-project.org/package=dplyr

15. Sievert C. plotly for R. Available: https://plotly-r.com

16. Garnier S. viridis: Default color maps from “matplotlib”. R package version 0.5.1. 2018. Available: https://CRAN.R-project.org/package=viridis

17. Hammer, Ø., Harper, D.A.T., Ryan, P.D. 2001. PAST: Paleontological statistics software package for education and data analysis. Palaeontologia Electronica 4(1): 9pp. http://palaeo-electronica.org/2001_1/past/issue1_01.htm

18. Grimm, E.C., 1991. Tilia and tiliagraph. Illinois State Museum, Springfield.

CHAPTER 4. LATE QUATERNARY VEGETATION, FIRE, AND HYDROCLIMATE AT THE SOUTHERN LIMIT OF THE TEMPERATE TALLGRASS PRAIRIE, MISSOURI, USA

Hannah M. Carrolla, Alan D. Wanamakerb, Lynn G. Clarka, Beth E. Caissieb, Jacqueline K.

Galangc, and Derek D. Houstond,1

aDepartment of Ecology, Evolution, and Organismal Biology, 2200 Osborn Drive, 251 Bessey

Hall, Iowa State University, Ames, Iowa, 50011, United States of America; bDepartment of

Geological and Atmospheric Sciences, 2237 Osborn Drive, 253 Science Hall I, Iowa State

University, Ames, Iowa, 50011, United States of America; cCollege of Liberal Arts and

Sciences, Iowa State University, Ames, Iowa, 50011, United States of America; dDepartment of

Natural Resource Ecology and Management, 2310 Pammel Drive, 339 Science Hall II, Ames,

Iowa, 50011, United States of America; 1Present address: Department of Natural and

Environmental Sciences, Western Colorado University, Hurst Hall 128, 1 Western Way,

Gunnison, CO 81231, United States of America

Modified from a manuscript to be submitted to Quaternary Science Reviews

Abstract

The boundaries (ecotones) between vegetation types are often highly sensitive to perturbations in climate. For this reason, ecotones have long been studied to reconstruct climate and vegetation dynamics. The northern prairie-forest ecotone has been extensively studied, but little is known about late Quaternary climate at its southern reaches. The tallgrass prairie-Ozark forest ecotone in southwest Missouri is an ecologically unique border between the temperate tallgrass prairie and warmer southern forests. We employed a multiproxy approach combining analyses of phytoliths (a direct proxy for past vegetation), charcoal (a proxy for paleofire), and stable carbon isotopes from soil (a proxy for the C3:C4 ratio of the vegetation which produced it),

71 to explore climate at the tallgrass prairie-Ozark forest ecotone over the last 19,000 years. We found that our site has remained open grassland for the span of the record, which is highly unusual for this region. Phytoliths record a mixed prairie for much of the mid to late-Holocene, with brief shifts to shortgrass prairie in times of water stress and a modern shift to tallgrass prairie that persists today. Encroachment by woody dicotyledons has been minimal. Fire was rare until the mid-Holocene, and frequent thereafter, which is consistent with the few other available records from the area. Stable carbon isotopes indicate generally high proportions of soil organic matter input from C4 plants, with relatively high sensitivity to hydroclimatic variability. We find that climate and vegetation in this region of southwest Missouri over most of the late Quaternary may have been impacted by a steeper east-west moisture gradient than exists today.

Introduction

Changes in climate are most quickly reflected in the vegetation at boundaries (ecotones) separating one ecoregion from another. These transitional areas may change rapidly and repeatedly as climate variations favor one vegetation type over another. For example, the prairie- forest ecotone in the eastern North American Great Plains has been extensively studied to reconstruct shifts in climate throughout the Holocene, because this ecotone is often highly responsive during such events (e.g., Baker et al., 2002; Dee & Palmer, 2017; Denniston et al.,

2000; Umbanhowar et al., 2006; Williams et al., 2009). Prairie advanced eastward across northern Iowa and southern Minnesota by more than 300 km (200 miles) in response to increased aridity from 10 kya – 9 kya (Williams et al. 2009). Prairie expansion took place at the expense of more mesic forests. Prairie receded westward again during periods of increased moisture, which favored expansion of forests (Denniston et al. 1999). High spatial heterogeneity and climatic sensitivity of prairie-forest mosaics points to the need for paleoclimatic data at high spatiotemporal resolutions. Focusing these efforts on ecotones improves the sensitivity of

72 paleoclimatic reconstructions, and of our understanding of the climate system as a whole. It also improves predictions of future vegetation responses under various global climate change scenarios. The ecotone separating prairie and forest at the southern limit of its range in Missouri and Oklahoma has been less extensively studied than prairie-forest ecotones further north (Jones et al., 2017; King, 1973), and many questions yet remain as to its provenance and stability.

The North American midcontinent has been dominated by grasslands since the late

Oligocene and early Miocene (Axelrod, 1985; Edwards et al., 2010; Strömberg, 2002, 2004,

2005). Changes in the ratio of C3 to C4 grasses have closely tracked climate, with periods of cooler, wetter conditions favoring C3 and either warmer or drier conditions favoring C4 (e.g.,

McInerney et al., 2011). At the Last Glacial Maximum, the area now occupied by Texas and

Oklahoma is estimated to have been 60 – 80% C4 grasses, with C3 grasses dominant further north and west across the midcontinent (Cotton et al., 2016).

Although the North American midcontinent has long been dominated by grasslands, modern vegetation assemblages are post-glacial in origin (Jackson and Williams, 2004). Prairies quickly expanded their range following the Last Glacial Maximum (LGM; approximately 22 kya in this region), and had largely moved into their present positions by the mid-Holocene (Baker et al.,

2000; Wright, 1992). The dominance of C4 grasses came to extend northward into Iowa and

Nebraska, with a roughly equal mix of C3 and C4 into Minnesota and South Dakota. North into

Canada and west to the Rocky Mountains, C3 grasses were strongly favored by the mid-

Holocene (Cotton et al., 2016), as is the case today.

Humans and megafauna have long inhabited the Great Plains region. Human activities are apparent in the archeological record in the form of land clearing, intentional and accidental fire, and the effects of hunting on the behavior of large mammals (e.g., Axelrod, 1985). Humans are

73 thought to have had substantial impacts on the distribution of North American grasslands since the onset of the Holocene. Bison are known to have been a keystone species throughout the

Great Plains until their extirpation by European settlers in the 90th Century (Allred et al., 2011;

Knapp et al., 1999). Bison increase the spatial heterogeneity of vegetation mosaics and can prevent the encroachment of trees into grasslands (Knapp et al., 1999). However, the large-scale effects of human-bison interactions on vegetation distributions are not well understood.

The modern-day temperate tallgrass prairie region extends from the Blackland Prairie sub- region of Texas into northeastern Oklahoma, and north to Saskatchewan, where a narrow band of prairie continues into southeastern Alberta (Figure 4-1a) (Commission for Environmental

Cooperation & Secretariat, 1997; Omernik, 1987; Risser et al., 1981). The eastern boundary of the temperate tallgrass prairie extends from Manitoba to western Minnesota, to the eastern borders of Iowa and Missouri, and is bordered by temperate forests for much of its range. The western boundary of the temperate tallgrass prairie is primarily bordered by mixed grass prairie.

At the southeastern end of its range in Missouri, it is flanked by the Ozark forest (Figure 4-1b).

Little is known of late Quaternary prairie-forest dynamics at the tallgrass prairie-Ozark forest ecotone. Few records exist, most of which are limited to the Holocene, with fewer still extending to the LGM. We sought to refine understanding of late Quaternary climate, vegetation, and fire in this critical transitional region between the temperate prairies of the north and the warm, humid forests of the south.

Much of our understanding of late Quaternary hydroclimate in the North American midcontinent is derived from pollen records, and to a lesser degree, macrofossils. These records are typically recovered from lake or streambank sediments and thus require suitable natural water bodies. Where natural water bodies are lacking, spatial coverage of paleoclimatic records

74 has suffered. This includes much of the western and southern Great Plains of North America. In the karstic region of southern Missouri, significant surface water is almost entirely absent except for artificial reservoirs. Pollen or macrofossils are therefore not a viable option for paleoclimate reconstructions.

We chose proxies that are stable in terrestrial soils to reconstruct paleoclimate and fire: phytoliths, charcoal, and stable carbon isotopes from soil organic matter. Silica phytoliths, also known as plant opal, are common microfossils produced by a wide variety of plants (Piperno,

2006). Phytoliths are taxonomically significant, readily identifiable, and stable in terrestrial soils over geologic time (Piperno, 1988). Phytoliths are reliable proxies for the plant communities from which they are derived when properly calibrated (e.g., Alexandre et al., 1997; Gao et al.,

2018; Hyland et al., 2013). Macroscopic charcoal (150 µm to > 250 µm) is a proxy for local to regional fires (e.g., Patterson et al., 1987; Scott, 2010; Touflan et al., 2010). In this range, particles are unlikely to have traveled long distances on wind. Soil carbon isotopes reflect the isotopic signature of the vegetation which produced the carbon (e.g., Davidson, 1995; Johnson et al., 2007; Kelly et al., 1998). They are commonly used to reconstruct the C3 to C4 ratio of the community, and therefore serve as a hydroclimate proxy. The combination of multiple environmental proxies can increase the robustness of paleoclimate reconstructions and provides checks against human error or contamination.

In this study, we employ a multiproxy approach using phytoliths, charcoal, and soil carbon isotopes to improve understanding of long-term vegetation, fire, and hydroclimate dynamics near the understudied, but ecologically important, temperate tallgrass prairie-Ozark forest ecotone.

We present a high-resolution record derived from a soil core taken in Golden Prairie, Missouri, near the location of the modern ecotone. We reconstruct past grassland type, woody dicotyledon

cover, water stress, fire frequency, and the relative contribution of C3 vs C4 vegetation to the soil organic carbon. We also use manganese oxide nodules to examine soil inundation and evaluate the potential for ferromanganous nodules to be used as sources of organic carbon for radiocarbon dating of terrestrial soil.

Methods

Study site

Golden Prairie (37.356°N, -94.149°W) is a 630 acre (254 ha) high quality temperate tallgrass prairie located near Golden City in southeast Barton County, Missouri, USA (Figure 4-1b). It is owned and managed by the Missouri Prairie Foundation (https://www.moprairie.org/). Of its 630 acres, 320 acres (129 ha) of Golden Prairie are intact remnant and the rest is reconstructed prairie

(Schuette, 2016). The remnant portion of the prairie, known as the Golden Prairie Natural Area, has been designated a National Natural Landmark and is part of the Missouri Natural Areas program.

Golden Prairie is situated in the Ozarks highlands, a physiographic region on the Ozarks plateau (Vineyard, 1997), approximately 2 miles (3.5 km) from the tallgrass prairie-Ozark forest ecotone as defined by the EPA’s Level II Ecoregions of North America (Commission for

Environmental Cooperation & Secretariat, 1997). The site consists mainly of flat to rolling upland with drainage to a small intermittent creek on the southern border of the property. The bedrock underlying Golden Prairie is Mississippian age (359 mya – 323 mya) high-calcium limestone (Vineyard, 1997). Patches of open oak savanna and pasturelands surround Golden

Prairie on all sides, although at the time of European settlement in the 1830s, it was reported that there were no forested areas nearby (Krusekopf & Bucher, 1914). Major Paleoindian sites are located 39 miles (63 km) northeast of Golden Prairie at Big Eddy.

Figure 4-1: a) Location of the temperate tallgrass prairie and Ozark forest. b) Golden Prairie is located approximately 2 miles (3.5 km) from the tallgrass prairie-Ozark forest ecotone in southwest Missouri. Black rectangle in the overview map indicates the extent of the inset map.

Soil core extraction and processing

Two soil cores were extracted from a shallow mesic swale in the upland remnant section of

Golden Prairie. LIDAR data available from the Missouri Spatial Data Information Service

(MSDIS) were used to select a coring location well away from major water drainage channels and the site of a previously undocumented historical farm plot. We recovered two intact soil cores, three inches (7.6 cm) in diameter, using a truck-mounted hydraulic coring device

(Giddings Machine Company, Windsor, CO, USA). Cores were taken approximately one meter apart. Both cores reached bedrock and returned the full soil profile of the site. The main core was

102 cm in length and contained a chip of limestone bedrock in the base. The backup core was

91.5 cm in length, and was archived under cold storage at Iowa State University following Initial

Core Description (ICD). Each core was split lengthwise and separated into a working half and an archival half. ICD was completed using a Munsell soil color chart (Munsell Color (firm), 2010) and visual description. The surface of the working half of the main core was scraped to reveal fresh soil, and then split horizontally at a 1 cm resolution. Each 1 cm section was placed into individual plastic bags. Remaining core halves were sealed, and all soil material was stored at

5°C.

Radiocarbon dating and age model creation

Phytoliths were extracted for radiocarbon dating using a much larger amount of soil than was necessary for counting (20 – 30 g). The same extraction procedure described in Section 2.5 was followed except that samples were agitated by hand only instead of being placed in a sonicating water bath. All glassware used in the process was combusted at 525°C for at least 8 hours, and plasticware was acid washed in 10% nitric acid for 48 hours prior to use. NanoPure™ water (Thermo Scientific, Waltham, MA, USA) was used in all steps after the clay removal series. The clean, washed phytolith sample was treated with a second hydrochloric acid wash and a second nitric acid wash, each followed by three water washes, to ensure removal of all contaminants prior to submission for radiocarbon dating. Water was used instead of acetone to transfer the final suspension to acid-washed vials, and samples were dried at 60°C for 24 – 48 hours. A small amount of material was mounted on glass slides prior to submission for radiocarbon dating.

Bulk soil and manganese oxide nodules used for radiocarbon dating were pretreated according to the acid rinse methods of Bao et al. (2019). Material was placed in precombusted glass test tubes and covered with approximately 10 mL of 1.0 N hydrochloric acid. Tubes were placed in a water bath at 60°C for one hour, centrifuged at 2,500 rpm, and the supernatant

78 decanted. Samples were rinsed with NanoPure™ water (Thermo Scientific, Waltham, MA,

USA). Manganese oxide nodules were additionally treated with 10% sodium hydroxide at 60°C for one hour to remove any humic acids, rinsed in NanoPure water, and decanted. Samples were dried at 60°C for at least 48 hours prior to being placed in precleaned tubes.

Phytoliths extracted for radiocarbon dating were dated using accelerator mass spectrometry at Beta Analytic, Inc., Miami, Florida. Bulk soil and manganese oxide nodules were dated using accelerator mass spectrometry by the Woods Hole Oceanographic Institution

National Ocean Sciences AMS Facility (NOSAMS), Woods Hole, MA. Raw radiocarbon dates were calibrated to IntCal13 (Reimer et al., 2013) and are reported in calibrated years before present (Cal BP) unless otherwise noted.

The proxy age model consists of 4 dates from manganese oxide nodules and 3 dates from phytoliths. The Bayesian age model was created from MnOX and phytolith dates using R package ‘rbacon’ ver. 2.4.0 (Blaauw et al., 2020).

Phytolith analysis

Phytolith extraction was completed using a modified version of Piperno (2006). 10 g (± 1g) of soil was weighed into 50 mL centrifuge tubes. Tubes were filled with a 5% solution of sodium hexametaphosphate and vortexed to break up soil aggregations. Particularly clay-rich samples were vortexed repeatedly over a period of one to three days as needed. Tubes were then placed in a sonicating water bath for 10 minutes after Lombardo et al. (2016), centrifuged for 10 minutes at 1500 rpm, and the supernatant decanted. Tubes were refilled with deionized water and vortexed to ensure that aggregates were sufficiently dispersed. Sonication was repeated a second time if necessary. Each sample was then rinsed by filling the tube to 50 mL with deionized water, vortexing, centrifuging at 1500 rpm for 10 minutes, and decanting the supernatant. This

79 was repeated three times to ensure complete removal of the sodium hexametaphosphate solution.

Rinsed samples were then wet sieved at 250 µm into 1000 mL beakers. The > 250 µm fraction was dried at 50°C and retained for analyses.

To remove clay, sieved samples were put through a series of gravity settles. Samples were placed in 1000 mL beakers. The beakers were filled to 1000 mL with deionized water using a strong flow to agitate the soil, and then allowed to settle for one hour. After one hour, the supernatant was carefully poured off and the beakers refilled with deionized water. The settling process was repeated until the supernatant was completely clear at the end of one hour; this typically required 8 – 10 (max 12) repetitions. Once clays were removed, soil samples were washed back into 50 mL centrifuge tubes, centrifuged, and decanted. Tubes were drained as thoroughly as possible before proceeding.

Carbonates were removed using 10 – 37% hydrochloric acid. Carbonate-poor samples were reacted with more dilute acid, while carbonate rich samples, the vast majority, were reacted using 37% HCl. Tubes were filled to 30 mL with acid, vortexed, and placed in a boiling water bath. Samples were periodically agitated, and fresh acid was added as needed. The process was allowed to continue until no further reaction was noted. Some samples required the spent acid to be removed and fresh acid added multiple times, or to be left overnight at room temperature in addition to several hours of boiling. Once carbonate removal was complete, samples were centrifuged and decanted, followed by one wash in deionized water.

Organic material was removed using nitric acid. Tubes were filled to 30 mL with concentrated nitric acid and placed in a boiling water bath. Small amounts of potassium chlorate were added to speed the reaction. Samples were occasionally agitated and more acid added as needed. The reaction was allowed to proceed until complete; usually several hours. Organic

80 removal was deemed sufficient when the supernatant turned a clear, deep yellow with no brown cast or cloudiness, and no new reaction was observed when fresh nitric acid and/or potassium chlorate was added. Once organic removal was complete, samples were centrifuged at 1500 rpm for 10 minutes and decanted, and then washed in deionized water four times.

Phytoliths were extracted via heavy liquid separation. A cadmium iodide/potassium iodide heavy liquid was prepared at a density of 2.3 g/mL. Tubes were filled to 10 mL with heavy liquid, vortexed, and centrifuged at 2500 rpm for 10 – 15 minutes. Phytoliths were skimmed from the surface of the tubes using Pasteur pipettes and transferred to clean 15 mL centrifuge tubes. This was repeated until no more phytoliths rose to the surface after centrifugation, usually 1 – 2 more times. The phytolith suspension was diluted with deionized water at a 2:1 to 3:1 ratio, vortexed, centrifuged at 2500 rpm for 10 minutes, and decanted. This was followed by 5 water washes to ensure complete removal of heavy liquid.

Following the final water wash, the prepared phytolith extraction was washed into glass vials with acetone and allowed to dry thoroughly in a fume hood prior to storage. A small amount of dry phytolith suspension was mounted on glass microscope slides in Permount™

(Fisher Chemical, Pittsburg, PA).

Phytoliths were identified using a reference collection prepared by the authors for

Missouri plants from specimens available in the Ada Hayden Herbarium (ISC) at Iowa State

University, Ames, Iowa, USA. Taxa were selected from Steyermark’s Flora of Missouri

(Yatskievych & Steyermark, 1999). Images of all phytolith reference specimens are available in the PhytCore Database (https://www.phytcore.org/). Additional reference was made to

Mulholland & Rapp (1992), Piperno (2006), and International Committee for Phytolith

Taxonomy (ICPT) et al. (2019) during identification. Phytolith nomenclature adheres to the

International Code for Phytolith Nomenclature (ICPN) 2.0 (ICPT et al., 2019).

Phytoliths were identified and counted on a Leica upright microscope (Leica Microsystems

GmbH, Wetzlar, Germany) or a Nikon Eclipse Ni-U (Nikon Instruments, Inc., Melville, NY) using differential interference contrast (DIC). A magnification of 400x or 600 – 1000x with oil immersion was used for identification. Vertical transects at random horizontal positions across the slide were counted until at least 200 – 300 diagnostic phytoliths were recorded. Counts stopped at the end of a transect if at least 200 diagnostic phytoliths had been recorded and no new diagnostic morphotypes had been recorded at that transect. Relative proportions of morphotypes were calculated from total phytolith count (diagnostic + nondiagnostic), excluding those that were broken, hidden, or too poorly preserved for identification. Counts of microscopic charcoal, diatoms, spheroid echinate sponge spicules, and the number of burned grass phytoliths were also recorded. See Table 1 for the diagnostic categories employed in this work.

Table 4-1: Diagnostic categories of phytoliths used to create vegetation indices. Naming conventions follow the International Committee for Phytolith Nomenclature (ICPN) 2.0 (International Committee for Phytolith Taxonomy (ICPT) et al., 2019). Grouping is given where a particular morphotype is indicative of a lower taxonomic resolution than the broad indicator category; all others are undifferentiated. In cell type, GSSC refers to the grass silica short cell category of Twiss et al. (1969), with names updated to the ICPN 2.0 taxonomy. Spheroid echinate morphotypes in temperate regions are produced by freshwater sponges and are included here as diagnostic silica after Yost et al. (2013).

Name (ICPN 2.0) Indicator Group Cell type Acute bulbosus Grasses (Poaceae) Undiff. Hair cell Bilobate Grasses (Poaceae) Panicoideae GSSC Bulliform flabellate Grasses (Poaceae) Undiff. Bulliform cell Conical Grasses (Poaceae) Pooideae GSSC Crenate Grasses (Poaceae) Pooideae/PACMAD GSSC Cross Grasses (Poaceae) Panicoideae GSSC Papillate, Cyperaceae Sedges (Cyperaceae) Undiff. Epidermal Papillate, non-Cyperaceae Grasses (Poaceae Pooideae GSSC Polylobate Grasses (Poaceae) Panicoideae/PACMAD GSSC Rondel Grasses (Poaceae) Pooideae GSSC Saddle Grasses (Poaceae) Chloridoideae GSSC

Table 4-1 (continued)

Name (ICPN 2.0) Indicator Group Cell type Spheroid echinate Freshwater sponge Undiff. Microsclere Spheroid ornate Woody dicots Undiff. Sclereid Trapezoid Grasses (Poaceae) Pooideae/PACMAD GSSC

The proportion of bulliform flabellate phytoliths relative to other grass phytolith morphotypes increases with increasing water stress (Bremond et al., 2005). This property serves to define the water stress index (Fs), which is expressed as percentage and calculated as:

퐵푢푙푙푙𝑖푓표푟푚 푓푙푎푏푒푙푙푎푡푒 퐹푠 = ∗ 100 퐺푟푎푠푠 푠𝑖푙𝑖푐푎 푠ℎ표푟푡 푐푒푙푙 푝ℎ푦푡표푙𝑖푡ℎ푠

Iph measures the balance between tallgrass and shortgrass prairie indicator taxa

(Alexandre et al., 1997). This index is expressed as a percentage and calculated as:

푆푎푑푑푙푒 퐼푝ℎ = ∗ 100 푆푎푑푑푙푒 + 퐶푟표푠푠 + 퐵𝑖푙표푏푎푡푒

An Iph index > 20% is indicative of shortgrass prairie, while Iph < 20% is indicative of tallgrass prairie. Values near 20 are interpreted to indicate mixed prairie.

Tree cover is estimated by the D/P ratio. We used the modified ratio of Bremond et al.

(2017), wherein only spheroid ornate (formerly globular granulate) phytoliths are included in the

D term, and the grass silica short cell phytolith (GSSC) category of Twiss et al. (1969) (Table 1) is used in the P term after Aleman et al. (2014) as follows:

퐷 퐷/푃 = 퐷 + 푃

A D/P of 0.1 is interpreted to be the crossover point between grassland and shrub-steppe, with values lower than 0.1 indicating prairie, and higher values indicating increasing woody dicotyledon cover.

Macroscopic charcoal analysis

Soil for macroscopic charcoal analysis was prepared following a modification of Horn &

Underwood (2014). 1 cm3 of soil (± 0.1 cm3) was measured via displacement of distilled water in a graduated cylinder. The soil was transferred to a 50 mL plastic centrifuge tube and the tube was filled with 5% sodium hexametaphosphate. Samples were agitated by hand and left to sit for 24 –

36 hours. Agitation was repeated as many times as necessary to completely deflocculate the soil.

Samples were then centrifuged at 1200 rpm for 5 minutes and the supernatant decanted. Tubes were refilled with distilled water and vortexed. Washed samples were wet sieved under distilled water at 250 µm, 150 µm, and 53 µm. Each size fraction was placed into an individual Petri dish and allowed to dry at 50°C for a minimum of 48 hours. Charcoal was counted under a lighted dissecting microscope at sufficient magnification to allow for positive identification of particles.

Manganese oxide nodules recovered in charcoal samples were counted simultaneously.

Charcoal is reported as the number of particles per cm3 of soil at each size fraction.

Stable carbon isotope methods

Methods for isotopic analysis of bulk soil were modified after Harris et al. (2001). Care was taken to remove any fragments of roots or other plant or animal matter. Approximately 2 g of soil was dried at 50°C for 24 hours and ground using a mortar and pestle. A 20 mg (± 1 mg) subsample was weighed into a silver capsule. One drop of water was added to each capsule, and the samples were placed in a sealed desiccator jar with a beaker containing 100 mL of 37% hydrochloric acid. Fumigation was allowed to proceed for 24 hours, after which samples were placed in a 50°C drying oven for 24 hours. Samples were then cooled to room temperature in a desiccator, reweighed, and the silver capsules sealed. Each silver capsule was then sealed into a tin capsule. Stable isotopes of carbon and nitrogen were measured in the Stable Isotopes Paleo

Environmental Research Group (SIPERG) laboratory at Iowa State University (Ames, IA, USA) on a Finnigan™ MAT Delta Plus XL mass spectrometer (Thermo Electron Corporation, San

Jose, CA, USA) in continuous flow mode connected to a Costech elemental analyzer (Costech

Analytical Technologies Inc., Valencia, CA, USA). 13C was corrected to Vienna Pee Dee

Belemnite (VPDB) and 15N was corrected to an air standard. Uncertainty is reported as the combined analytical uncertainty and average correction factor.

Statistical analyses were performed in R version 3.6.2 (R Core Team, 2019). To calculate the proportion of organic carbon contributed by C3 versus C4 plants, we used the mass balance equation:

13 13 13 훿 퐶 = (훿 퐶퐶4)(푥) + (훿 퐶퐶3)(1 − 푥) where 13C is the measured value produced by isotopic analysis of soil, expressed in permil (‰),

13 13 13  CC4 is the global average value of  C measured in C4 plants (-13‰);  CC3 is the global

13 average value of  C measured in C3 plants (-27‰), and x is the proportion of soil organic carbon contributed by C4 plants, after Derner et al. (2006). Kohn (2010) recommends using a

13 global estimate  C for C3 plants of -28.5‰ instead of -27‰ due to spatial bias in the estimation of the latter; we report both. The estimates of soil organic matter contributed by C4 when -27‰

13 is used for the  CC3 term are reported in the text, followed by estimates produced by using a

13  CC3 of -28.5‰ in parentheses. In both cases, we multiplied the resultant proportion by 100, and the estimated amount of soil organic carbon contributed by C4 plants is expressed as a percentage unless otherwise indicated.

X-ray diffraction

X-ray diffraction was used to determine the crystalline structure of manganese oxide

(MnOX) nodules. Nodules were prepared for x-ray diffraction in the X-Ray Diffraction

Laboratory at Iowa State University. Clean, dry MnOX nodules were ground to a fine, uniform powder in a mortar and pestle. A small amount of material was packed into a slide well and measured on a Rigaku Ultima IV (Rigaku Corporation, Tokyo, Japan). Output was processed in

JADE (Materials Data, Inc., Livermore, California).

Results

Age model

Bayesian age modeling based on MnOX and phytolith radiocarbon dates produced a mean soil accumulation rate of 200 years/cm for the proxy record (Figure 4-2). The maximum median age at the base of the core (101 – 102 cm) is 19,204 Cal BP with a range of 17,337 Cal BP – 21,406

Cal BP. The median age estimate at the top of the core (0 – 1 cm) is 116 Cal BP with a range of -

66 Cal BP to 846 Cal BP. The age model is nearly linear and produces a modeled age at the maximum depth in the core markedly younger than the radiocarbon date of 26,446 ± 2,127 Cal

BP at 75 cm.

Figure 4-2: Bayesian age model for the proxy record from manganese oxide and phytolith dates.

Phytoliths

Phytoliths occur in high enough numbers for quantitative analyses from the surface of the core to a depth of ~63 cm (~11.7 kya). Phytoliths are rare from 63-80 cm (~11.7 – 14.9 kya), and entirely absent from the record below 80 cm. Partial dissolution is evident beginning at ~20 cm

(~3.7 kya), but dissolution is not severe enough to render analysis impossible until ~64 cm

(~11.8 kya).

Overall, GSSC phytoliths (as defined by Twiss et al. 1969) comprise from 2.5% to 21%

(average 11.4%) of all identifiable phytoliths (diagnostic + nondiagnostic) throughout the quantifiable portion of the core, and are present at every counted depth. The rondel type is the most common GSSC phytolith type, accounting for 1 – 13% (average 7%) of total phytoliths.

This is followed by trapezoid (0.5 – 2.9%, average 1.3%), bilobate (0 – 2.8%, average 0.8%), and crenate (0 – 2.4%, average 0.8%). The remaining types, conical, cross, polylobate, saddle, and non-Cyperaceae papillate comprise, on average, <0.5% of the total phytolith count.

Occurrence of other grass forms is highly variable. Acute types (trichomes) range from

2.4 – 12.4% of the total. Bulliform flabellate morphotype percentages range from 0.2 – 11.8%

(average 4.2%). These types are produced by all grasses and cannot be resolved further (Piperno,

2006).

Non-grass morphotypes are generally rare. The spheroid ornate morphotype is typically associated with cover of woody dicots (Alexandre et al., 1997). Relative abundance is low throughout the core, ranging from 0 – 0.9% (average 0.4%). The Cyperaceae morphotype ranges from 0 – 2.1% (average 0.7%).

Other biogenic silica

The spheroid echinate morphotype in temperate soils is indicative of freshwater sponge microsclere production, a proxy for standing water (Uriz et al., 2003). Relative abundance is generally low, ranging from 0 – 3.1% (average 0.4%) The highest proportion occurs at depth, where few phytoliths remain.

Diatoms occur only infrequently, with no more than 1 individual pennate diatom (2 valves) present in the phytolith counts except at 4 cm (~0.8 kya) (18 valves), and are unlikely to be of use as proxy. We therefore exclude them from further discussion.

Phytolith indices

The water stress index, Fs, indicates moderately high water stress between 62 cm (~11.5 kya) and 50 cm (~8.5 kya) (Figure 4-6). Its maximum of 44% was reached at 58 cm (~10.5 kya).

Fs gradually declined until reaching its minimum of 1% at 20 cm (~3.6 kya). Fs increased gradually throughout the late Holocene before declining to 6% at 4 cm (~0.75 kya).

Phytolith index lph measures the balance between tallgrass and shortgrass indicators. The

88 lph index reveals that Golden Prairie has oscillated between shortgrass (Iph >20) and mixed grass prairie (Iph near 20) multiple times, with tallgrass prairie (Iph <20) only becoming dominant in the last ~4 kya. Only shortgrass indicator morphotypes were present at 50 cm and the Iph is accordingly 100%. This is coincident with the second highest water stress peak. Too few GSSC morphotypes were found at the depth of the maximum water stress peak (58 cm) to calculate Iph, which hampers interpretation.

The D/P ratio is used to measure relative woody dicotyledon cover. At present, Golden

Prairie is entirely open grassland, with no tree cover near enough to contribute phytoliths. The modern D/P ratio is 0. D/P is low throughout the record, with slight tree cover signals present prior to the mid-Holocene. At 54 cm (~ 9.3 kya), the D/P ratio of 0.15 indicates a short-lived transition to shrub-steppe.

Charcoal

Macroscopic charcoal analysis indicates that fire was rare from the base of the record until between 39 cm and 32 cm (~6.5 kya – ~5.6 kya). Minor fire pulses are apparent at 90 cm,

92 cm, 64 cm, and 59 cm to 44 cm. Both frequency and apparent intensity of fires increased sharply after 39 cm, peaking at 17 cm (~3 kya) with 418 particles/cm3 of soil. Charcoal then declined until reaching 11 particles/cm3 at 13 cm (~2.2 kya), and peaked again at 9 cm (~ 1.5 kya) with 281 particles/cm3, and at 3 cm (~0.6 kya) with 354 particles/cm3. We were unable to sample from 5 – 7 cm and 1 – 3 cm due to insufficient material, and missing data are not to be interpreted as a signal of hiatus.

Microscopic charcoal is absent from much of the record, with minor pulses occurring at

42 cm, 38 cm, 30 cm, 20 cm, 18 cm, 14 cm, and 4 cm. Burnt grass phytoliths appear at 62 cm, 18 cm, and 4 cm, and range from 0.6% to 1.3% of grass phytoliths evidently having been burned.

Soil organic carbon

Analyses of stable carbon isotopes from bulk soil indicate that generally stable conditions persisted from the base of the record (~19 kya) to approximately 79 cm (~15 kya). Mixed C3 and

C4 plants were present during this period, with C4 grasses contributing, on average, 78% (80% if

13 a  CC3 of -28.5‰ is used; see Methods) of the soil organic carbon (SOC). Rapid climatic change follows the end of this relatively stable period. Between 79 cm (~15 kya) and 77 cm

(~14.3 kya), climatic oscillation is evident in a 13C decrease from -15.29‰ to -16.23‰, which corresponds to an increase in the SOC contribution by C4 grasses from approximately 77%

(79%) to 84% (85%). Between 77 cm and 71 cm (~14 kya – 13 kya), a series of rapid

13 oscillations terminated in a decrease in  C of 1.53‰, equivalent to a reduction in C4 SOC contribution from 84% (85%) at 77cm to 75% (77%) at 71 cm. A period of conditions more favorable to C4 followed at 70 cm (~12.8 kya) and persisted until 64 cm (~11.8 kya), during which time the SOC contribution by C4 plants rose to 88% (89%). The SOC contribution by C4 then decreased to 79% (81%) by 59 cm (~10.9 kya), corresponding to a 13C of -16.0‰. 13C then increased to its maximum value in the record, -14.4‰ by 53 cm (~9 kya), indicating that C4 plants contributed 90% (91%) of SOC at this time. Between 53 cm and 34 cm (~5.7 kya) remained generally high, from 83% – 91% (84% –92%), with an average of 88% (89%), corresponding to an average 13C of -14.7‰. A slow decline in 13C followed, reaching -15.3‰,

13 equivalent to 84% (85%) C4 SOM contribution, at 11 cm (~1.8 kya). A rapid decrease in  C of approximately 2.7‰ occurred in the top ~10 cm of the core, reaching -18.0‰ at 0 cm.

We found low organic carbon percentages in bulk soil from the base of the record to approximately 40 cm, averaging 0.55% (range 0.3% – 0.85%). Above ~40 cm organic carbon percentages increased gradually to a maximum of 4.35% in the top 1 cm of soil, with the steepest increase beginning at 4 cm.

Manganese oxide nodules

X-ray diffraction analysis confirmed that the manganese oxide nodules were amorphous

(Figure 4-3). The only significant peak occurred between 26 and 27 degrees, which is consistent with quartz inclusions from soil.

Figure 4-3: Output from x-ray diffraction on powdered manganese oxide nodules. The only appreciable peak was identified as quartz, likely contamination from soil.

Manganese oxide nodules are present throughout the core, and their highest concentrations occur above 59 cm (Fig 4-6). Manganese oxide nodule concentration peaks at 57 cm (598 particles/cm3) and 55 cm (607 particles/cm3) and remains generally higher in the top half of the core than in the bottom half.

Figure 4-4: Soil carbon isotopic record and diagnostic silica types found at Golden Prairie by age (based on Bayesian age modeling). The spheroid echinate morphotype is interpreted as a signal of freshwater sponge presence. Spheroid ornate is indicative of woody dicotyledons. Grass silica short cells (GSSC) follow the conventions of Twiss et al. (1969). Acute bulbosus and bulliform flabellate types are non-specific grass indicators. Error bars for 13C are combined analytical and instrumental uncertainty. Silica morphotypes are reported as relative abundance.

Figure 4-5: Soil carbon isotopic record, generic grass indicator morphotype relative abundance (proportions), and grass silica short cell morphotype relative abundance (proportions) by indicator group. Bulliform flabellate and acute bulbosus are widely produced in the grasses and cannot be resolved further. GSSC types are understood to degrade more quickly than other morphotypes in situ (Piperno 2006) and therefore caution should be used in interpreting the oldest assemblages. Bilobate and polylobate (combined due to rarity of the latter) and cross are diagnostic of the Panicoideae; conical, crenate, rondel, and trapezoid are indicative of the Pooideae; saddles are produced by the Chloridoideae. Error bars for 13C are combined analytical and instrumental uncertainty.

Figure 4-6: Overview of main indicator groups found at Golden Prairie. Items in shades of purple are derived from soil inclusions 13 (charcoal, microcharcoal, manganese oxide (MnOX); items in shades of green are derived from soil organic matter ( C/%C4, % organic carbon); items in shades of teal are derived from phytolith counts (% burnt grass phytoliths, D/P, Fs, Iph). Dashed gray line in 13 13  C/%C4 subplot represents a linear projection of the trend in  C/%C4 in the absence of an appreciable Suess Effect. Dotted gray line in the D/P subplot is the dividing line between grassland and shrub-steppe. Error bars for 13C are combined analytical and instrumental uncertainty.

Discussion

We sought to refine understanding of late Quaternary fire, vegetation, and climate at the ecologically important, yet understudied tallgrass prairie-Ozark forest ecotone. The scarcity of natural lakes in this region has prevented the extensive pollen and macrofossil work which has taken place further north in the midcontinent, and which has served to provide a highly detailed picture of late Quaternary climate. We therefore employed markers of climate that are stable in terrestrial soils over geologic time to create a high-resolution record from a soil core taken in

Golden Prairie, Missouri.

We find strong evidence of open mixed C3/C4 grassland with little invasion of forest at

Golden Prairie throughout the last ~19,000 years. The 13C signature from bulk soil indicates contribution of SOC by C4 plants of at least 80% throughout most portions of the record, although this contribution should not be interpreted as implying percent cover or percent of biomass. Under favorable conditions, the C4 photosynthetic pathway is capable of producing more total organic carbon than C3 (Ehleringer et al., 1997). Actual C4 biomass proportion is therefore likely much lower than an estimate of 80% SOC contribution would seem to indicate.

Low pCO2 during the late Pleistocene yielded an advantage to C4 plants (Ehleringer et al., 1997), and it is possible that such an advantage might have remained even under somewhat unfavorable light conditions, e.g., shading by trees. However, had significant tree cover existed at Golden

Prairie during the late Pleistocene, its presence should be reflected in a much lower soil organic carbon 13C than we observed. Summer relative humidity in the midcontinent at the Last Glacial

Maximum has been reconstructed at 59%; much lower than the modern value approaching 68%

(Voelker et al., 2015). Low pCO2 combined with summer aridity would strongly favor C4 grasses, particularly if trees were absent. Phytolith dissolution in the lower 40 cm of the core

95 precludes a more direct estimate of plant community composition, but late Pleistocene 13C values in our core likely point to a mixed C3/C4 grassland without substantial local tree cover.

Our findings of open mixed grassland are highly unusual for this region in the late

Pleistocene. Nearby sites north and east of Golden Prairie are known to have been dominated by boreal spruce forest from ~20 – 13.5 kya (King, 1973). Spruce forest has also been confirmed in northeastern Kansas (Grüger, 1973), and through much of the midcontinent north to the margin of the Laurentide ice sheet (e.g., Williams et al., 2004). Mixed spruce and pine have been found in records from Cupola pond in southeastern Missouri during the late Pleistocene, although temperate tree species such as Quercus were certainly also present (Jones et al., 2017). We tentatively interpret our findings of a more open, mixed grassland to be a localized phenomenon in the absence of additional information.

Our reconstruction of infrequent fire activity throughout the late Pleistocene is consistent with several other records from across the region, and with a global synthesis of charcoal records spanning the late Quaternary by Power et al. (2008). Sustained cool conditions with rare sources of ignition (lightning) during the late glacial and early interglacial are generally thought to be the drivers.

We find evidence of climate instability at approximately 77 cm and 64 cm (14.3 kya –

11.8 kya) consistent with signals of abrupt changes in vegetation composition. Signals of abrupt change in vegetation at the Pleistocene-Holocene transition occur in other parts of southwest

Missouri and northwestern Arkansas, largely driven by changes in moisture (Williams et al.,

2009). Prairie is known to have advanced in Missouri approximately 8.5 kya (Baker & Waln,

1985), roughly coincident with our reconstruction of the onset of frequent fire and a shift from shortgrass to mixed grass prairie.

Tree cover estimates from the D/P ratio indicate that woody dicotyledon cover peaked around 54 cm (~9.2 kya) at about 0.15, indicative of shrub-steppe or savanna. D/P of 0.1 is recognized as a dividing line between shrub-steppe and grassland, with values of 1 or higher indicating closed forest, and values less than 0.1 indicating primarily open habitat (Alexandre et al., 1997; Strömberg, 2002). Decline in woody cover rapidly followed, coincident with the highest water stress (Fs) signal in the record at 50 cm (~8.5 kya) and a shortgrass index (Iph) of

100%. Our reconstruction of mid-Holocene aridity is consistent with records from across much of the midcontinent (Baker et al., 1992; Baker et al., 2002; Bartlein et al., 1984; Umbanhowar et al., 2006; Williams et al., 2009). Weak fire signals in the microscopic record during the mid-

Holocene indicate both woody biomass burning and occasional grass fire, but relatively little macroscopic charcoal was found. This may point to causes other than fire in the decline of shrub- steppe. D/P gradually declined after ~8.5 kya, never again approaching the 0.1 transition to shrub-steppe, even with highly variable, albeit lower, water stress.

Mixed grass prairie dominated Golden Prairie for much of our record. Iph is relatively steady at intermediate values, indicative of mixed grass prairie, between ~6.3 kya and 3.6 kya.

Fire became frequent throughout this period, corroborating the apparent transition to a fire- dependent prairie following the MCO termination. D/P values dropped to 0 by ~2.5 kya, coincident with an increase in water stress, indicating that Golden Prairie has been an entirely open grassland throughout the late Holocene. Tallgrass prairie dominated the system between

~3.3 kya and 2.3 kya, at the same time that fire frequency reached is maximum, and thereafter returned to a more mixed prairie type. A gradual decline in 13C during the same period points to a moderate late Holocene climate. In the top 11 cm of the core, approximately the depth of the O horizon, we find strong evidence of the 13C Suess Effect (Keeling et al., 2017; Suess, 1955) The

modern-day prairie is a C4-dominated tallgrass system (Institute of Botanical Training, LLC,

2014), and more work is needed to determine the timing of its return.

Phytolith dissolution in the lower 40 cm of our core complicates interpretation of our record. Phytolith preservation in the top ~63 cm of the core is sufficient to enable direct estimates of vegetation dynamics. Dissolution does become appreciable near this limit, however, and it is possible that short cells, which are known to dissolve before other morphotypes

(Piperno, 2006), are underrepresented. We therefore stress a cautious interpretation of our findings near the zone of dissolution.

We find a strong signal of continuous grassland dominance at Golden Prairie since at least 19 kya, with little evidence of invasion of woody dicotyledons despite the presence of forest at nearby sites since the late Pleistocene (e.g., King, 1973). There are no physical barriers to prevent invasion of forest at Golden Prairie. Fire in the vicinity of Golden Prairie was infrequent until well into the mid-Holocene, so it is unlikely to have been a key mechanism in the maintenance of the grassland during the late Pleistocene and early to mid-Holocene. The late

Pleistocene vegetation of the Ozarks, near which Golden Prairie is located, has been described as a patchy mosaic of vegetation types with a sharp east-west gradient in climate (Kay, 2012).

Golden Prairie climate at the end of the Pleistocene appears to have been more akin to that further south and west than to the rest of the Ozark highlands. Much additional work will be needed to ascribe a specific cause.

Extensive archaeological evidence shows that humans have inhabited this region for at least the last 13,000 years (Lopinot et al., 1998; Ray et al., 1998). It is therefore probable that local vegetation and fire records contain a signature of human habitation. Between approximately 450 CE and the time of European expansion, the Osage Nation continuously

98 inhabited much of southern Missouri and were certainly employing intentional burns for management purposes (Nanavati & Grimm, 2019). In addition, the presence of bison in the region could have exerted strong controls on the expansion of forest into the prairie at our site.

We found no direct evidence of bison or other large mammals in our soil core, but cannot rule out their influence at Golden Prairie. More work is needed to disentangle the climate signal from possible anthropogenic and megafaunal signals.

The tallgrass prairie-Ozark forest ecotone serves as an excellent model system, as it represents a transitional zone between temperate grasslands and warm forests. Major climatic transitions in the late Pleistocene and Holocene would be expected to destabilize the grassland- forest boundary and allow for the advance of forest. However, we found strong evidence that

Golden Prairie has remained grassland, with limited woody dicotyledon cover, for at least the last 19 kyr. Other nearby regions were forested in the late Pleistocene and remain so today, but we found no definitive signals of significant forest encroachment.

MnOX nodules are common worldwide in both terrestrial and marine settings (Post,

1999), and we successfully radiocarbon dated soil that was trapped within some of them.

Evaluation of manganese oxide nodules by x-ray diffraction confirmed that their structure is amorphous, which aids in confirming their identity (Post, 1999). Radiocarbon dating of the soil trapped within nodules at our site has been limited in scope, but shows promise as a potential method for obtaining radiocarbon dates in terrestrial soils. No prior attempt has been made to employ the nodules for radiometric dating. Future work should pair manganese oxide nodule dating with direct dating of phytoliths or other macroscopic plant remains to test the hypothesis that manganese oxide nodules can serve as a radiocarbon dating tool for terrestrial soils comparable to dating of macroscopic and/or microscopic plant remains.

Golden Prairie lies close to both modern and past ecoregion boundaries. In the late

Pleistocene, this portion of southwestern Missouri may have experienced climate more akin to the warmer conditions modeled by Cotton et al. (2016) in northeastern Oklahoma and southwest

Kansas than the cooler conditions recorded immediately to the east. Recent modeling work by

Lora & Ibarra (2019) provides evidence that the present-day Golden Prairie may have been near a critical boundary between relatively warm and moist Gulf air masses and cooler northern air masses throughout the late Pleistocene. Additional work is needed to understand whether the apparently anomalous climate signal recorded in our core was more widespread at the modern- day tallgrass prairie-Ozark forest ecotone than has been previously recognized, or is purely a local phenomenon.

Data availability

Upon acceptance for publication, all data will be made publicly available through the

Neotoma Paleoecology Database and this statement will be updated to include pertinent dataset and site identification numbers, a DOI, and a permanent link to the dataset.

References

Aleman, J. C., Canal-Subitani, S., Favier, C., & Bremond, L. (2014). Influence of the local environment on lacustrine sedimentary phytolith records. Palaeogeography, Palaeoclimatology, Palaeoecology, 414, 273–283. https://doi.org/10.1016/j.palaeo.2014.08.030

Alexandre, A., Meunier, J.-D., Lézine, A.-M., Vincens, A., & Schwartz, D. (1997). Phytoliths: indicators of grassland dynamics during the late Holocene in intertropical Africa. Palaeogeography, Palaeoclimatology, Palaeoecology, 136(1–4), 213–229. https://doi.org/10.1016/S0031-0182(97)00089-8

Allred, B. W., Fuhlendorf, S. D., & Hamilton, R. G. (2011). The role of herbivores in Great Plains conservation: comparative ecology of bison and cattle. Ecosphere, 2(3), 1-17.

100

Axelrod, D. I. (1985). Rise of the grassland biome, central North America. The Botanical Review, 51(2), 163–201.

Baker, R. G., & Waln, K. A. (1985). Quaternary pollen records from the Great Plains and central United States. In R. G. Holloway & V. M. Bryant (Eds.), Pollen records of late- quaternary North American sediments (pp. 191–203). Dallas: American Association of Stratigraphic Palynologists Foundation.

Baker, R. G., Maher, L. J., Chumbley, C. A., & Van Zant, K. L. (1992). Patterns of Holocene environmental change in the midwestern United States. Quaternary Research, 37(3), 379–389. https://doi.org/10.1016/0033-5894(92)90074-S

Baker, R.G., Bettis, E. A., Denniston, R. F., Gonzalez, L. A., Strickland, L. E., & Krieg, J. R. (2002). Holocene paleoenvironments in southeastern Minnesota – chasing the prairie- forest ecotone. Palaeogeography, Palaeoclimatology, Palaeoecology, 177(1–2), 103– 122. https://doi.org/10.1016/S0031-0182(01)00354-6

Baker, R. G, Fredlund, G. G., Mandel, R. D., & Bettis, E. A. I. (2000). Holocene environments of the central Great Plains: multi-proxy evidence from alluvial sequences, southeastern Nebraska. Quaternary International, 67(1), 75–88.

Bao, R., McNichol, A. P., Hemingway, J. D., Lardie Gaylord, M. C., & Eglinton, T. I. (2019). Influence of different acid treatments on the radiocarbon content spectrum of sedimentary organic matter determined by RPO/accelerator mass spectrometry. Radiocarbon, 61(2), 395–413. https://doi.org/10.1017/RDC.2018.125

Bartlein, P. J., Webb, T., & Fleri, E. (1984). Holocene climatic change in the northern Midwest: Pollen-derived estimates. Quaternary Research, 22(3), 361–374. https://doi.org/10.1016/0033-5894(84)90029-2

Blaauw, M., Christen, J. A., & Aquino, M. A. (2020). rbacon: Age-depth modelling using Bayesian statistics. (Version 2.4.0) [R package]. Retrieved from https://CRAN.R- project.org/package=rbacon

Bremond, L., Alexandre, A., Peyron, O., & Guiot, J. (2005). Grass water stress estimated from phytoliths in West Africa: Grass water stress estimated from phytoliths. Journal of Biogeography, 32(2), 311–327. https://doi.org/10.1111/j.1365-2699.2004.01162.x

Bremond, L., Bodin, S. C., Bentaleb, I., Favier, C., & Canal, S. (2017). Past tree cover of the Congo Basin recovered by phytoliths and δ13C along soil profiles. Quaternary International, 434, 91–101. https://doi.org/10.1016/j.quaint.2015.11.150

Commission for Environmental Cooperation, & Secretariat. (1997). Ecological regions of North America: toward a common perspective. Montréal, Québec: The Commission.

Cotton, J. M., Cerling, T. E., Hoppe, K. A., Mosier, T. M., & Still, C. J. (2016). Climate, CO2, and the history of North American grasses since the Last Glacial Maximum. Science Advances, 2(3), e1501346. https://doi.org/10.1126/sciadv.1501346

101

Davidson, G. R. (1995). The stable isotopic composition and measurement of carbon in soil CO2. Geochimica et Cosmochimica Acta, 59(12), 2485–2489. https://doi.org/10.1016/0016- 7037(95)00143-3

Dee, J. R., & Palmer, M. W. (2017). Annual rings of perennial forbs and mature oaks show similar effects of climate but inconsistent responses to fire in the North American prairie– forest ecotone. Canadian Journal of Forest Research, 47(6), 716–726. https://doi.org/10.1139/cjfr-2016-0473

Denniston, R. F., González, L. A., Baker, R. G., Asmerom, Y., Reagan, M. K., Edwards, R. L., & Alexander, E. C. (1999). Speleothem evidence for Holocene fluctuations of the prairie- forest ecotone, north-central USA. The Holocene, 9(6), 671–676.

Denniston, R. F., González, L. A., Asmerom, Y., Reagan, M. K., & Recelli-Snyder, H. (2000). Speleothem carbon isotopic records of Holocene environments in the Ozark Highlands, USA. Quaternary International, 67(1), 21–27. https://doi.org/10.1016/S1040- 6182(00)00005-7

Derner, J. D., Boutton, T. W., & Briske, D. D. (2006). Grazing and ecosystem carbon storage in the North American Great Plains. Plant and Soil, 280(1–2), 77–90. https://doi.org/10.1007/s11104-005-2554-3

Edwards, E. J., Osborne, C. P., Stromberg, C. A. E., Smith, S. A., Bond, W. J., Christin, P. A., et al. (2010). The origins of C4 grasslands: Integrating evolutionary and ecosystem science. Science, 328(587), 587–591. https://doi.org/10.1126/science.1177216

Ehleringer, J. R., Cerling, T. E., & Helliker, B. R. (1997). C4 photosynthesis, atmospheric CO2, and climate. Oecologia, 112(3), 285–299. https://doi.org/10.1007/s004420050311

Gao, G., Jie, D., Liu, L., Liu, H., Li, D., Li, N., et al. (2018). Assessment and calibration of representational bias in soil phytolith assemblages in Northeast China and its implications for paleovegetation reconstruction. Quaternary Research, 90(1), 38–49. https://doi.org/10.1017/qua.2018.5

Grüger, J. (1973). Studies on the late Quaternary vegetation history of northeastern Kansas. Geological Society of America Bulletin, 84(1), 239–250.

Harris, D., Horwáth, W. R., & van Kessel, C. (2001). Acid fumigation of soils to remove carbonates prior to total organic carbon or CARBON-13 isotopic analysis. Soil Science Society of America Journal, 65(6), 1853. https://doi.org/10.2136/sssaj2001.1853

Horn, S. P., & Underwood, C. A. (2014). Methods for the study of soil charcoal as an indicator of fire and forest history in the Appalachian region, U.S.A., 8.

Hyland, E., Smith, S. Y., & Sheldon, N. D. (2013). Representational bias in phytoliths from modern soils of central North America: Implications for paleovegetation reconstructions. Palaeogeography, Palaeoclimatology, Palaeoecology, 374, 338–348. https://doi.org/10.1016/j.palaeo.2013.01.026

102

Institute of Botanical Training, LLC. (2014). Golden Prairie floristic integrity report (p. 43 pp). Salem, Missouri, USA: Missouri Prairie Foundation.

International Committee for Phytolith Taxonomy (ICPT), Neumann, K., Strömberg, C. A. E., Ball, T., Albert, R. M., Vrydaghs, L., & Cummings, L. S. (2019). International code for phytolith nomenclature (ICPN) 2.0. Annals of Botany, 124(2), 189–199. https://doi.org/10.1093/aob/mcz064

Johnson, W. C., Willey, K. L., & Macpherson, G. L. (2007). Carbon isotope variation in modern soils of the tallgrass prairie: Analogues for the interpretation of isotopic records derived from paleosols. Quaternary International, 162–163, 3–20. https://doi.org/10.1016/j.quaint.2006.10.036

Jones, R. A., Williams, J. W., & Jackson, S. T. (2017). Vegetation history since the last glacial maximum in the Ozark highlands (USA): A new record from Cupola Pond, Missouri. Quaternary Science Reviews, 170, 174–187. https://doi.org/10.1016/j.quascirev.2017.06.024

Kay, M. (2012). The Ozark highland Paleoarchaic. In From the Pleistocene to the Holocene: Human Organization and Cultural Transformations in Prehistoric North America (Vol. 17, p. 233). Texas A&M University Press.

Keeling, R. F., Graven, H. D., Welp, L. R., Resplandy, L., Bi, J., Piper, S. C., et al. (2017). Atmospheric evidence for a global secular increase in carbon isotopic discrimination of land photosynthesis. Proceedings of the National Academy of Sciences, 114(39), 10361– 10366. https://doi.org/10.1073/pnas.1619240114

Kelly, E. F., Blecker, S. W., Yonker, C. M., Olson, C. G., Wohl, E. E., & Todd, L. C. (1998). Stable isotope composition of soil organic matter and phytoliths as paleoenvironmental indicators. Geoderma, 82(1–3), 59–81. https://doi.org/10.1016/S0016-7061(97)00097-9

King, J. E. (1973). Late Pleistocene palynology and biogeography of the western Missouri Ozarks. Ecological Monographs, 43(4), 539–565. https://doi.org/10.2307/1942305

Knapp, A. K., Blair, J. M., Briggs, J. M., Collins, S. L., Hartnett, D. C., Johnson, L. C., & Towne, E. G. (1999). The keystone role of bison in North American tallgrass prairie: Bison increase habitat heterogeneity and alter a broad array of plant, community, and ecosystem processes. BioScience, 49(1), 39-50.

Kohn, M. J. (2010). Carbon isotope compositions of terrestrial C3 plants as indicators of (paleo)ecology and (paleo)climate. Proceedings of the National Academy of Sciences, 107(46), 19691–19695. https://doi.org/10.1073/pnas.1004933107

Krusekopf, H. H., & Bucher, F. S. (1914). Soil survey of Barton County, Missouri (p. 32). Washington, D.C.: U.S. Department of Agriculture.

103

Lombardo, U., Ruiz-Pérez, J., & Madella, M. (2016). Sonication improves the efficiency, efficacy and safety of phytolith extraction. Review of Palaeobotany and Palynology, 235, 1–5. https://doi.org/10.1016/j.revpalbo.2016.09.008

Lopinot, N. H., Ray, J. H., & Conner, M. D. (1998). The 1997 Excavations at the Big Eddy Site (23CE426) in Southwest Missouri: Fort Belvoir, VA: Defense Technical Information Center. https://doi.org/10.21236/ADA373957

Lora, J. M., & Ibarra, D. E. (2019). The North American hydrologic cycle through the last deglaciation. Quaternary Science Reviews, 226, 105991. https://doi.org/10.1016/j.quascirev.2019.105991

McInerney, F. A., Strömberg, C. A. E., & White, J. W. C. (2011). The Neogene transition from C 3 to C 4 grasslands in North America: stable carbon isotope ratios of fossil phytoliths. Paleobiology, 37(1), 23–49. https://doi.org/10.1666/09068.1

Mulholland, S. C., & Rapp, G. (Eds.). (1992). Phytolith Systematics: Emerging Issues. Springer US. https://doi.org/10.1007/978-1-4899-1155-1

Munsell Color (firm). (2010). Munsell soil color charts: with genuine Munsell color chips. Grand Rapids, MI.

Nanavati, W. P., & Grimm, E. C. (2019). Humans, fire, and ecology in the southern Missouri Ozarks, USA. The Holocene, 095968361987580. https://doi.org/10.1177/0959683619875807

Omernik, J. M. (1987). Ecoregions of the Conterminous United States. Annals of the Association of American Geographers, 77(1), 118–125. https://doi.org/10.1111/j.1467- 8306.1987.tb00149.x

Patterson, W. A., Edwards, K. J., & Maguire, D. J. (1987). Microscopic charcoal as a fossil indicator of fire. Quaternary Science Reviews, 6(1), 3–23. https://doi.org/10.1016/0277- 3791(87)90012-6

Piperno, D. R. (1988). Phytolith analysis: An archaeological and geological perspective. San Diego, CA: Academic Press.

Piperno, D. R. (2006). Phytoliths: a comprehensive guide for archaeologists and paleoecologists. Rowman Altamira.

Post, J. E. (1999). Manganese oxide minerals: Crystal structures and economic and environmental significance. Proceedings of the National Academy of Sciences, 96(7), 3447–3454. https://doi.org/10.1073/pnas.96.7.3447

Power, M. J., Marlon, J., Ortiz, N., Bartlein, P. J., Harrison, S. P., Mayle, F. E., et al. (2008). Changes in fire regimes since the Last Glacial Maximum: an assessment based on a global synthesis and analysis of charcoal data. Climate Dynamics, 30(7–8), 887–907. https://doi.org/10.1007/s00382-007-0334-x

104

R Core Team. (2019). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from https://www.R-project.org/

Ray, J. H., Lopinot, N. H., Hajic, E. R., & Mandel, R. D. (1998). The Big Eddy Site: A multicomponent Paleoindian site on the Ozark Border, Southwest Missouri. Plains Anthropologist, 43(163), 73–81. https://doi.org/10.1080/2052546.1998.11931884

Reimer, P. J., Bard, E., Bayliss, A., Beck, J. W., Blackwell, P. G., Ramsey, C. B., et al. (2013). IntCal13 and Marine13 radiocarbon age calibration curves 0–50,000 years Cal BP. Radiocarbon, 55(4), 1869–1887.

Risser, P. G., Birney, E. C., & Blocker, H. D. (1981). The true prairie ecosystem.

Schuette, B. (2016). 15. The conservation significance of prairie remnants in Missouri. In North American Prairie Conference Proceedings (Vol. 4, pp. 171–180).

Scott, A. C. (2010). Charcoal recognition, taphonomy and uses in palaeoenvironmental analysis. Palaeogeography, Palaeoclimatology, Palaeoecology, 291(1–2), 11–39. https://doi.org/10.1016/j.palaeo.2009.12.012

Strömberg, C. A. E. (2002). The origin and spread of grass-dominated ecosystems in the late Tertiary of North America: preliminary results concerning the evolution of hypsodonty. Palaeogeography, Palaeoclimatology, Palaeoecology, 177(1–2), 59–75. https://doi.org/10.1016/S0031-0182(01)00352-2

Strömberg, C. A. E. (2004). Using phytolith assemblages to reconstruct the origin and spread of grass-dominated habitats in the great plains of North America during the late Eocene to early Miocene. Palaeogeography, Palaeoclimatology, Palaeoecology, 207(3–4), 239– 275. https://doi.org/10.1016/j.palaeo.2003.09.028

Strömberg, C. A. E. (2005). Decoupled taxonomic radiation and ecological expansion of open- habitat grasses in the Cenozoic of North America. Proceedings of the National Academy of Sciences, 102(34), 11980–11984. https://doi.org/10.1073/pnas.0505700102

Suess, H. E. (1955). Radiocarbon concentration in modern wood. Science, New Series, 122(3166), 415–417.

Touflan, P., Talon, B., & Walsh, K. (2010). Soil charcoal analysis: a reliable tool for spatially precise studies of past forest dynamics: a case study in the French southern Alps. The Holocene, 20(1), 45–52. https://doi.org/10.1177/0959683609348900

Twiss, P. C., Suess, E., & Smith, R. M. (1969). Morphological classification of grass phytoliths 1. Soil Science Society of America Journal, 33(1), 109. https://doi.org/10.2136/sssaj1969.03615995003300010030x

105

Umbanhowar, C. E., Camill, P., Geiss, C. E., & Teed, R. (2006). Asymmetric vegetation responses to mid-Holocene aridity at the prairie–forest ecotone in south-central Minnesota. Quaternary Research, 66(1), 53–66. https://doi.org/10.1016/j.yqres.2006.03.005

Uriz, M.-J., Turon, X., Becerro, M. A., & Agell, G. (2003). Siliceous spicules and skeleton frameworks in sponges: Origin, diversity, ultrastructural patterns, and biological functions. Microscopy Research and Technique, 62(4), 279–299. https://doi.org/10.1002/jemt.10395

Vineyard, J. D. (1997). An introduction to: The geology of Missouri. Rocks & Minerals, 72(6), 363–369. https://doi.org/10.1080/00357529709605066

Voelker, S. L., Stambaugh, M. C., Guyette, R. P., Feng, X., Grimley, D. A., Leavitt, S. W., et al. (2015). Deglacial hydroclimate of midcontinental North America. Quaternary Research, 83(2), 336–344. https://doi.org/10.1016/j.yqres.2015.01.001

Williams, J. W., Shuman, B. N., Webb, T., Bartlein, P. J., & Leduc, P. L. (2004). Late- Quaternary vegetation dynamics in North America: Scaling from taxa to biomes. Ecological Monographs, 74(2), 309–334. https://doi.org/10.1890/02-4045

Williams, J. W., Shuman, B., & Bartlein, P. J. (2009). Rapid responses of the prairie-forest ecotone to early Holocene aridity in mid-continental North America. Global and Planetary Change, 66(3–4), 195–207. https://doi.org/10.1016/j.gloplacha.2008.10.012

Wright, H. E. (1992). Patterns of Holocene climatic change in the midwestern United States. Quaternary Research, 38(1), 129–134.

Yatskievych, G. A., & Steyermark, J. A. (1999). Steyermark’s flora of Missouri. Missouri Department of Conservation.

Yost, C. L., Blinnikov, M. S., & Julius, M. L. (2013). Detecting ancient wild rice (Zizania spp. L.) using phytoliths: a taphonomic study of modern wild rice in Minnesota (USA) lake sediments. Journal of Paleolimnology; Dordrecht, 49(2), 221–236. http://dx.doi.org.proxy.lib.iastate.edu/10.1007/s10933-012-9670-x

Acknowledgements

Funding for this work was provided by two Geological Society of America Graduate

Student Research Grants, three awards of the Iowa State University Department of Ecology,

Evolution, and Organismal Biology Harry and Audrey Finch Scholarship, and one award of the

Iowa State University Department of Ecology, Evolution, and Organismal Biology Lois H.

106

Tiffany Scholarship. The authors thank Richard Baker for advice regarding site selection and laboratory methods. We thank E. Arthur Bettis III for advice regarding coring device selection and for providing laboratory supplies for phytolith extractions. We thank Aaron Sassman for providing training for use of the hydraulic coring device and associated equipment. We thank

Arkady Ellern for training and support in x-ray diffraction. We are grateful to Deborah Lewis for access to specimens in the Ada Hayden Herbarium and permission for destructive sampling, and

Kathryn Holmes for assistance in preparing the phytolith reference collection. We thank Suzanne

Ankerstjerne for laboratory analysis of stable isotope samples. We thank Rosa Maria Albert and

José Antonio Ruiz García of the PhytCore database (https://www.phytcore.org/) for their assistance with reference collection image uploads. This work was done in cooperation with the

Missouri Prairie Foundation (https://www.moprairie.org/). We are especially grateful to Carol

Davit, Bruce Schuette, and Jerod Huebner for access to Golden Prairie and permission to extract soil cores.

107

CHAPTER 5. GENERAL CONCLUSIONS

The goals of this dissertation are to examine patterns of vegetation and climate dynamics at regional to subcontinental scales across the North American midcontinent throughout the late

Quaternary and provide new tools for analysis. Patterns of vegetation distribution at regional to subcontinental scales are the result of a complex suite of interactions between biotic and abiotic factors working over a multiplicity of spatiotemporal scales. I was inspired by the synthesis of

Seddon et al. (2014) to consider the common signals I might be able to find in paleorecords from across the North American midcontinent. Pollen ratios have long been employed to finely discriminate between past vegetation types (Herzschuh 2007, Toth et al. 2011, Zhao et al. 2012), generally with great local success but little broad applicability. I expanded and refined the

Ambrosia to Artemisia ratio of Morris (2013) and discovered that not only could a common signal be found in the ratio’s response to precipitation across the midcontinent, it also had wider applicability. I hope that this tool will prove useful for future researchers in defining past ecoregion boundaries and reconstructing past climate.

I spent a great deal of time exploring pollen data available through the Neotoma

Paleoecology Database (Williams et al. 2018) at a subcontinental scale to understand what pressing gaps in the paleoecological record that my skillset might aid in filling. One of those gaps was the accessibility of pollen data to a broad audience without a coding background. To fill that gap, I developed GeoPollen, the application that would have been most useful to me early on in my graduate career when I was trying to understand the spatiotemporal distribution of vegetation over the late Quaternary, and also the spatiotemporal distribution of the pollen records themselves. The ability to quickly and easily visualize complex data is immensely helpful when seeking to understand broad patterns. My hope for GeoPollen is for it to be an ongoing labor of

108 love for me, a project that grows and evolves through time. A consequence of exploring so much data was a recognition of areas where more data was needed. My study of vegetation and climate in Golden Prairie, Missouri was undertaken because I saw a critical gap in the paleoecological record. Few studies of late Quaternary climate have been conducted in the area (e.g., King 1973,

Jones et al. 2017), not for lack of interest on the part of the research community, but because most common paleoproxies do not survive in terrestrial soils. I chose to employ alternative proxies which do survive, and I found a complex vegetation and fire history that requires more fine-scale analyses to fully understand. Much more work is needed to tease apart complex regional patterns of climate and vegetation at the tallgrass prairie-Ozark forest ecotone, and evaluate the role of human and megafaunal activity there. However, I found that vegetation and climate at Golden Prairie were akin to more arid sites to the south and west than to those immediately east and north. This agrees with previous work that found evidence of a steep east- west moisture gradient across the midcontinent which no longer exists. My dissertation work has served to identify common environmental signals at multiple spatiotemporal scales, particularly in the form of hydroclimatic signals as recorded by floral paleoproxies, and to create new tools for their analysis. Across the North American midcontinent, I found widespread signals of high spatiotemporal heterogeneity, which we can use to improve our understanding of climate at ever- higher resolutions.

References

Herzschuh, U. 2007. Reliability of pollen ratios for environmental reconstructions on the Tibetan Plateau. Journal of Biogeography 34:1265–1273.

Jones, R. A., J. W. Williams, and S. T. Jackson. 2017. Vegetation history since the last glacial maximum in the Ozark highlands (USA): A new record from Cupola Pond, Missouri. Quaternary Science Reviews 170:174–187.

109

King, J. E. 1973. Late Pleistocene Palynology and Biogeography of the Western Missouri Ozarks. Ecological Monographs 43:539–565.

Morris, C. J. 2013. Analysis of modern pollen data from the prairies of central North America. Thesis, Kansas State University, Manhattan, Kansas.

Toth, I., R. Peternel, L. Srnec, and B. Vojnikovi. 2011. Diurnal Variation in Airborne Pollen Concentrations of the Selected Taxa in Zagreb, Croatia. Coll. Antropol.:8.

Zhao, Y., H. Liu, F. Li, X. Huang, J. Sun, W. Zhao, U. Herzschuh, and Y. Tang. 2012. Application and limitations of the Artemisia /Chenopodiaceae pollen ratio in arid and semi-arid China. The Holocene 22:1385–1392.

110

APPENDIX A. GEOPOLLEN SOURCE CODE AND DOCUMENTATION

This appendix gives all source code for the release version of GeoPollen (Version 1.0).

For the most current source code, see GeoPollen’s GitHub repository

(https://github.com/hannahcarroll/GeoPollen). This source code was written using package

‘shinydashboard’ version 0.7.1 in R version 3.6.2.

The global.R script contains code which runs automatically at app launch. This code performs the necessary steps to pre-process and clean data ahead of users accessing it. The ui.R script builds the interface that users will see and interact with. The server.R script defines variables, contains all functions available to users, and handles downloads. The script get_downloads.R is a standalone file which downloads data from Neotoma each month and readies it to be pulled into GeoPollen’s Docker container. It is called from within the global.R script. It runs silently and is not seen by users. Each user session is created using the server.R and ui.R scripts. The two supporting scripts are R Markdown documents. These are userguide.Rmd, which provides a brief guide on how to use each aspect of GeoPollen, and citeneotoma.R, which provides information to users on how to cite the resources they have used for analyses within GeoPollen.

Global.R

# Define list of required packages packages <- c("shiny", "shinydashboard", "leaflet", "dplyr", "xtable", "analogue", "magrittr",

"rgeos", "rgdal", "viridis", "sf", "tidyr", "sp", "shinyjs", "DT", "ggplot2", "plotly", "data.table",

"neotoma", "cronR")

# Load packages lapply(packages, require, character.only = TRUE)

111

########################################################################

# Automatically download data monthly

# This will only work on Linux#

# Clear any old jobs cron_clear(ask=FALSE)

# Get filepath f <- system.file(package = "cronR", "directorywithinDocker", "get_downloads.R")

# Define path cmd <- cron_rscript(f)

# Create the job cron_add(cmd, frequency = 'monthly', id = 'monthlyupdate', at = '00:05', days_of_month = 'first', days_of_week = '*')

# End

##############################################################################

# Data rawdata <- read.csv("nam whitmore full 22k.csv") rawdata <- rawdata %>%

rename(

lng = long

)

# Remove varve dated records datasets <- rawdata[rawdata$date.type != "Varve years BP", ] rm(rawdata) # Remove raw data from memory

112

# Remove spikes and non-pollen palynomorphs datasets <- datasets[ , -which(names(datasets) %in% c("LYCOPODX", "EQUISETU",

"SPHAGNUM", "PTERIDIUM"))] datasets[, 11:120][is.na(datasets[, 11:120])] <- 0 # Recode NAs to 0 in the pollen columns

# Calibrate radiocarbon ages correctionfactor <- read.table("intcal0914c.txt", sep=",") names(correctionfactor) <- c("calBP", "age"," Error", "D14Cpermil", "sigma14C")

# Reimer et al. (2009) correction.curve <- lm(calBP ~ age, data=correctionfactor)

# Apply correction factor

# Copy the age column to a new column datasets$calendar.ybp <-datasets$age

# Split out the ages that are already calibrated splpol.cal <- within(datasets,

{calendar.ybp[date.type=="Calendar years BP"] <- calendar.ybp[date.type=="Calendar years

BP"];

calendar.ybp[date.type=="Calibrated radiocarbon years BP"] <- calendar.ybp[date.type=="Calibrated radiocarbon years BP"]})

# Grab the data with raw radiocarbon dates

113 splitpollen.ybp <- subset(datasets, date.type=="Radiocarbon years BP")

# Use the correction curve to predict the calibrated age splitpollen.ybp$calendar.ybp <- predict(correction.curve, splitpollen.ybp)

# Put them back together datasets2 <- rbind(splpol.cal, splitpollen.ybp)

# Round the ages to the nearest year datasets2$calendar.ybp <- as.integer(datasets2$calendar.ybp)

# Cleanup rm(correction.curve) rm(correctionfactor) rm(splpol.cal) rm(splitpollen.ybp)

# Pollen sums datasets3 <- datasets2 %>%

mutate(sumVar = rowSums(.[11:120]))

# Drop all zero rows datasets3 <- filter(datasets3, sumVar != 0)

114

# Calculate proportions datasets4 <- datasets3[,(11:120)] %<>% sapply(`/`, datasets3[,122])

# Recombine into one datasets5 <- cbind(datasets3[,c(1:10,121)], datasets4)

# Cleanup rm(datasets) rm(datasets2) rm(datasets3) rm(datasets4)

# Take just those 22k and younger datasets5 <- subset(datasets5[datasets5$calendar.ybp <= 22000, ])

# Make sure none of the data got duplicated datasets5 <- datasets5[!duplicated(datasets5[,c('dataset','depth')]),]

# Create unique rownames rownames(datasets5) <- paste(datasets5$site.name, datasets5$dataset, datasets5$depth, datasets5$calendar.ybp)

115

# Round pollen proportions datasets5[,12:121] <- round(datasets5[,12:121], 3)

# Round lat long datasets5[,8:9] <- round(datasets5[,8:9],6)

# Remove any with lat and long reported as NA datasets5 <- filter(datasets5, !is.na(lat) & !is.na(lng))

# Create a column to join on datasets5$unique <- rownames(datasets5)

# Setup for subsetting by state/province and region political <- readOGR("./NA_political", 'ne_50m_admin_1_states_provinces_lakes', encoding='UTF-8') political <- spTransform(political, CRS("+init=epsg:3857")) # Match Leaflet's default

# Point in poly to assign ecoregions to datasets

d.sp <- datasets5[,c(9,8)]

# Create the spatial object coordinates(d.sp) <- ~ lng + lat

116 proj4string(d.sp) <- CRS("+init=epsg:4326") # Native projection d.sp <- spTransform(d.sp, CRS("+init=epsg:3857")) # Match Leaflet's default

# Create the spatial polygons data frame

# The datsets get the state/province they are on top of

d.by.region <- over(d.sp, political)

# Create a field to join on d.by.region$unique <- rownames(d.by.region)

# Get just the columns we need d.by.region <- select(d.by.region, "region", "name", "unique") rm(d.sp)

# Merge dataset location with region datasets6 <- as.data.frame(merge(datasets5, d.by.region, by="unique"))

# Drop column we joined on datasets6 <- select(datasets6, -"unique")

# Cleanup rm(datasets5)

117 rm(d.by.region)

# Set the rownames again rownames(datasets6) <- paste(datasets6$site.name, datasets6$dataset, datasets6$depth, datasets6$calendar.ybp)

# Setup for the Distribution tab long_DF <- datasets6 %>% gather(Taxon, Count, ABIES:LARREA)

long_DF <- long_DF[long_DF$Count != 0, ] # Keep nonzero counts only long_DF <- long_DF[!is.na(long_DF$Taxon),] # Drop anything with taxa as NA (as a backup!) long_DF <- rename(long_DF, "Relative Proportion" = "Count") # We calculated the proportions earlier long_DF$Taxon <- as.factor(long_DF$Taxon) # Make it a factor to be safe

##############################################################################

Ui.R

# Define UI for application

header <- dashboardHeader(title = "GeoPollen")

body <- dashboardBody(

tags$head(tags$style(HTML('

.form-group, .selectize-control {

118

margin-bottom: 0px;

}

.form-group, .selectize-control {

margin-top: 0px;

}

.form-group, .selectize-control {

margin-left: 3px;

}

.form-group, .selectize-control {

margin-right: 3px;

}

.box-body {

padding-bottom: 3px;

}

.box-body {

padding-top: 2px;

}

.box {margin: 1px;}'))), useShinyjs(),

# Tabs tabItems(

tabItem(tabName = "Main",

119

fluidRow(

box(width = 9,

title = NULL, solidHeader = FALSE,

column(12,

tags$style(type = "text/css", "#map {height: calc(100vh - 275px) !important;}"),

leafletOutput("mainmap")

)

box(width = 3,

column(12,

title = "Controls",

sliderInput("agerange",

"Calendar Years Before Present",

min = 1950-(as.integer(format(Sys.Date(), "%Y"))),

max = max(20000),

value = c(8000, 10000)),

selectInput('regions', 'Search by Region (Optional)', levels(droplevels(datasets6$region)), multiple = TRUE),

selectInput('state', "Search by State/Province (Optional)", levels(droplevels(datasets6$name)), multiple = TRUE),

120

selectInput('site', 'Select Site by Name (Optional)', levels(droplevels(datasets6$site.name)), multiple = TRUE),

div(style="display:inline-block;width:49%;text-align: center;", actionButton("resetall", label =

"Reset All")),

div(style="display:inline-block",title = "Download ALL data from sites currently in view on the map",

downloadButton("d1", label = HTML("Download Sites
Currenly in View")))

)

fluidRow(

box(

title = NULL, width = 3, solidHeader = TRUE,

plotOutput("agehist", height = "200px")),

box(

title = "Summary (Visible Datasets)", width = 4, solidHeader = TRUE,

tableOutput("quicksummary")),

box(

title = "Analog Controls", width = 5, solidHeader = TRUE,

column(6,

121

div(style="display:inline-block;width:175px;text-align: center;",

tags$div(style="display:inline-block", selectInput("selectsite", "Select Age(s) to Use as a Reference (CalBP)",

choices = NULL, multiple = TRUE, width = "175px")),

tags$div(style="display:inline-block", actionButton("findanalogs", label = "Find

Analog Communities", title="Warning: Can be slow for large datasets", icon = icon("search", lib

= "font-awesome"))),

tags$div(style="display:inline-block", textOutput("confirmation"))

)),

column(6,

plotOutput("plot", height = "200px", width = "175px"))

)

tabItem(tabName = "analogstab"),

tabItem(tabName = "mindiss",

fluidRow(

box(width = 12, title = "Your training set is:",

tableOutput("trainset"))

fluidRow(

box(width = 12,

verbatimTextOutput("analogfun"))

122

)),

tabItem(tabName = "traindiag",

fluidRow(

box(width = 5,

plotOutput("traindiags")),

box(width = 7,

title = "Heatmap of Training Set Squared Chord Distance",

plotlyOutput("trainheat")))),

tabItem(tabName = "analagous",

fluidRow(

box(width = 12,

title = "Heatmap of Training vs Test Set Squared Chord Distance",

column(12,

tags$style(type = "text/css", "#testheat {height: calc(100vh - 50px)

!important;}"),

plotlyOutput("testheat")))),

fluidRow(

box(width = 12,

dataTableOutput("andf")

)

)),

#Distribution mapping tab

123

tabItem(tabName = "distmapping",

fluidRow(

box(width = 9,

title = NULL, solidHeader = FALSE,

column(12,

tags$style(type = "text/css", "#map {height: calc(100vh - 275px) !important;}"),

leafletOutput("distmap")

)),

box(width = 3,

column(12,

title = "Controls",

sliderInput("agerange2",

"Calendar Years Before Present",

min = 1950-(as.integer(format(Sys.Date(), "%Y"))),

max = max(20000),

value = c(8000, 10000)),

selectInput('taxa', 'Select Taxa (One or more)', long_DF$Taxon, multiple = TRUE),

div(style="display:inline-block;width:49%;text-align: center;", actionButton("resetall2", label

= "Reset ALL Selections")),

tags$div(style="display:inline-block",title = "Download ALL records for sites currently visible on map",

124

downloadButton("d2", label = HTML("Download Sites
Currently in View"))),

tags$div(style="display:inline-block",title = "Download ALL records for ALL sites reporting selected taxa",

downloadButton("d3", label = HTML("Download All
Mapped Sites")))

)),

box(width = 3,

column(12,

"Select taxa using dropdown menu or by typing part of the name"

))),

fluidRow(

box(width = 6, title = "Distribution (Total Range in Selected Time Period)", solidHeader = TRUE,

column(12,

tableOutput("quicksummary2"))),

box(width = 6, title = "Distribution (Currently in View)", solidHeader = TRUE,

column(12,

tableOutput("quicksummary3")))

fluidRow(

box(width = 12,

title = NULL, solidHeader = FALSE,

column(12,

125

h4(

a("Interpret past plant distributions with caution. There is known spatial bias in

Neotoma. For more information, see Inman et al. 2018", target = "_blank", href =

"https://www.sciencedirect.com/science/article/abs/pii/S0277379118303962")))))),

#Raw data tab

tabItem(tabName = "rawmain",

fluidRow(

box(width = 12,

column(12,

title = "Raw Data", solidHeader = TRUE,

dataTableOutput("rawdata")))

)),

tabItem(tabName = "rawdist",

fluidRow(

box(width = 12,

column(12,

title = "Raw Data", solidHeader = TRUE,

dataTableOutput("rawdata2"))))),

#Citation tab

tabItem(tabName = "cite",

fluidRow(

126

column(12,

includeMarkdown("citeneotoma.Rmd")))

tabItem(tabName = "about",

fluidRow(

column(12,

includeMarkdown("userguide.Rmd"))))

)

############ Dashboard Sidebar ############ sidebar <- dashboardSidebar(width = 200, sidebarMenu(

menuItem(

"Main", tabName = "Main", icon = icon("map", lib = "font-awesome")),

menuItem("Analog Communities", tabName = "analogstab", icon =icon("equals", lib = "font- awesome"),

menuSubItem("Minimum Dissimilarity", tabName = "mindiss"),

menuSubItem("Training Set Diagnostics", tabName = "traindiag"),

menuSubItem("Analogs", tabName = "analagous")),

127

menuItem("Distribution Mapping", tabName = "distmapping", icon = icon("map-marked-alt", lib = "font-awesome")

menuItem("Raw Data", tabName = "rawdatatab", icon = icon("file-alt", lib = "font-awesome"),

menuSubItem("Main Tab", tabName = "rawmain"),

menuSubItem("Distribution Mapping Tab", tabName = "rawdist")),

menuItem("Citations", tabName = "cite", icon = icon("align-right", lib = "font-awesome")

menuItem("Contact", icon = icon("address-card", lib = "font-awesome"),

a(actionButton(inputId = "email1", label = "Email",

icon = icon("envelope", lib = "font-awesome")),

href="mailto:[email protected]"),

a(actionButton(inputId = "twitter", label = "Twitter",

icon = icon("twitter", lib = "font-awesome"),

onclick ="window.open('https://twitter.com/hmcarro', '_blank')")),

a(actionButton(inputId = "github", label = "GitHub",

icon = icon("github", lib = "font-awesome"),

onclick ="window.open('https://github.com/hannahcarroll', '_blank')")),

a(actionButton(inputId = "bugs", label = "Submit a Bug Report",

icon = icon("bug", lib = "font-awesome"),

128

onclick ="window.open('https://github.com/hannahcarroll/GeoPollen/issues', '_blank')"))),

menuItem("How to Use GeoPollen", tabName = "about", icon = icon("question-circle", lib =

"font-awesome")),

menuItem(a("Go to the Neotoma", br(), "Paleoecology Database", target="_blank", href="https://www.neotomadb.org/"))

))

########### Dashboard page #################### dashboardPage(

header,

sidebar,

body

)

##############################################################################

Server.R

# Define server logic server <- function(input, output, session) {

# Basemaps

# Main tab map

output$mainmap <- renderLeaflet({

129

leaflet() %>%

addProviderTiles(providers$CartoDB.Positron) %>%

setView(lng=-96, lat=41, zoom=3.5)

})

output$distmap <- renderLeaflet({

leaflet() %>%

addProviderTiles(providers$CartoDB.Positron) %>%

setView(lng=-96, lat=41, zoom=3.5)

})

###### Main tab ######

# A reactive expression that returns the datasets that are in bounds right now

visibledatasets <- reactive({

if (is.null(input$site) & is.null(input$regions) & is.null(input$state)) {

return({

d2 <- subset(datasets6,

calendar.ybp >= input$agerange[1] &

calendar.ybp <= input$agerange[2]

)

bounds <- input$mainmap_bounds

latRng <- range(bounds$north, bounds$south)

130

lngRng <- range(bounds$east, bounds$west)

subset(d2,

lat >= latRng[1] & lat <= latRng[2] &

lng >= lngRng[1] & lng <= lngRng[2]

)

}

)

} else {

return({userselected()

})

}

})

# React to optional search parameters userselected <- reactive({

if(!is.null(input$site)) {

d3 <- datasets6 %>% dplyr::filter(site.name %in% paste(input$site))

}

if(!is.null(input$regions)) {

d3 <- datasets6 %>% dplyr::filter(region %in% paste(input$regions))

}

if(!is.null(input$state)) {

131

d3 <- datasets6 %>% dplyr::filter(name %in% paste(input$state))

}

subset(d3,

calendar.ybp >= input$agerange[1] &

calendar.ybp <= input$agerange[2]

)

})

# Histogram of visible dataset ages

output$agehist <- renderPlot({

if (nrow(visibledatasets()) == 0)

return(NULL)

hist(visibledatasets()$calendar.ybp,

main = "Visible Dataset Ages",

xlab = "Calendar ybp",

xlim = range(visibledatasets()$calendar.ybp),

ylab = "Record Count",

col = '#1B2631',

border = 'white'

)

})

132

# Summary of visible datasets

qsummary <- reactive({

visibledatasets() %>%

summarise("Age Old" = format(max(calendar.ybp), digits = 0, scientific=FALSE),

"Age Young" = format(min(calendar.ybp), digits = 0, scientific=FALSE),

"Record Count" = length(visibledatasets()$.id),

"Unique Sites" = length(unique(visibledatasets()$site.name))

)

})

output$quicksummary <- renderTable({

if (nrow(visibledatasets()) == 0)

return(NULL)

qsummary()

striped = TRUE,

rownames = FALSE,

spacing = "s",

align = "c"

)

# Optional inputs

133

observeEvent(input$regions,{

updateSelectInput(session, "regions", "Search by Region (Optional)",

choices = sort(unique(droplevels(datasets6$region))),

selected = c(input$regions))

})

observeEvent(input$state,{

updateSelectInput(session, "state", "Search by State/Province (Optional)",

choices = sort(unique(droplevels(datasets6$name))),

selected = c(input$state))

})

observeEvent(input$site,{

updateSelectInput(session, "site", "Select Site by Name (Optional)",

choices = sort(droplevels(datasets6$site.name)),

selected = c(input$site))

})

###

# Map display responds to user and markers are redrawn as necessary

observe({

if(is.null(input$site) & is.null(input$regions) & is.null(input$state)) {

leafletProxy("mainmap",

data = visibledatasets()) %>%

134

clearMarkers() %>%

addCircleMarkers(~lng, ~lat,

popup = paste(

visibledatasets()$site.name, "
",

"Lat:", format(visibledatasets()$lat, digits = 4, scientific=FALSE),

"Long:", format(visibledatasets()$lng, digits = 4, scientific=FALSE), "
",

"Dataset ID:", visibledatasets()$dataset),

label = NULL)

} else {

# Allows for selection of datasets by optional search parameters

leafletProxy("mainmap") %>%

clearMarkers() %>%

addCircleMarkers(data = userselected(), ~lng, ~lat,

popup = paste(

userselected()$site.name, "
",

"Lat:", format(userselected()$lat, digits = 4, scientific=FALSE),

"Long:", format(userselected()$lng, digits = 4, scientific=FALSE), "
",

"Dataset ID:", userselected()$dataset),

label = NULL)

}

})

135

# Analogs

# Start with site selection

observe({

selectedsite <- reactiveValues(clickedMarker=NULL)

observeEvent(input$mainmap_marker_click,{

selectedsite$clickedMarker <- input$mainmap_marker_click

refsite <- selectedsite$clickedMarker

refsite2 <- visibledatasets()[which(visibledatasets()$lat == refsite$lat & visibledatasets()$lng == refsite$lng), ]

updateSelectInput(session, "selectsite",

choices = sort(refsite2$calendar.ybp), # update choices

selected = NULL) # remove selection

output$plot=renderPlot({

if (is.null(input$mainmap_marker_click)) {

textOutput("Select a site on the map")

} else {

ggplot(refsite2) + geom_point(aes(x=round(calendar.ybp, 0), y=round(depth, 2))) +

136

labs(title = paste(refsite2$site.name),

subtitle = "(selected age range only)",

x = "CalBP", y = "Depth (cm)") + theme_bw() + scale_y_reverse()

}

})

# This is the user's training set trainingset <- reactive({

subset(refsite2, calendar.ybp %in% input$selectsite)

})

# Use action button to find analogs observeEvent( input$findanalogs, {

## merge training and test set on columns trainingset2 <- trainingset() %>% dplyr::select(AMBROSIA:ALRUBRA) testset <- visibledatasets() %>% dplyr::select(AMBROSIA:ALRUBRA)

#myanalogs <- join(trainingset2, testset, verbose = TRUE) analogs <- analog(trainingset2, testset, method = "SQchord", keep.train = TRUE)

#print(summary(analog(trainingset, testset, method = "SQchord"))) output$analogfun <- renderPrint({ analogs })

137

output$confirmation <- renderText({ "Diagnostics have been returned to the Analog

Communities tab" })

output$trainset <- renderTable({

trainingset() %>% dplyr::select(site.name:calendar.ybp)

})

mydissims <- dissim(analogs)

output$traindiags <- renderPlot(plot(mydissims))

output$trainheat <- renderPlotly({

m1 <- matrix(analogs$train, nrow = nrow(analogs$train), ncol = ncol(analogs$train))

p1 <- plot_ly(

x = c(rownames(analogs$train)), y = colnames(analogs$train),

z = round(m1, 3), type = "heatmap", zmin = 0, zmax = 2

)

return(p1)

})

output$testheat <- renderPlotly({

asdf <- as.data.frame(analogs$analogs)

m2 <- as.matrix(asdf, nrow = nrow(asdf), ncol = ncol(asdf))

p2 <- plot_ly(

x = c(colnames(asdf)), y = c(rownames(asdf)),

138

z = round(m2, 3), type = "heatmap", zmin = 0, zmax = 2

)

return(p2)

})

output$andf <- renderDataTable({

as.data.frame(analogs$analogs)%>%

mutate_if(is.numeric, round, digits = 3)

rownames = rownames(analogs$analogs),

options = list(paging=FALSE, scrollX = TRUE)

)

})

##### Distributions Tab #####

distributions <- reactive({

taxaInput <- input$taxa

selectedtaxa <- long_DF[long_DF$Taxon %in% taxaInput, ]

selectedtaxa2 <- subset(selectedtaxa,

calendar.ybp >= input$agerange2[1] &

139

calendar.ybp <= input$agerange2[2]

)

return(selectedtaxa2)

})

# Datasets currently in view

visibledist <- reactive({

if (is.null(input$distmap_bounds))

return(distributions()[FALSE,])

dist2 <- subset(distributions(),

calendar.ybp >= input$agerange2[1] &

calendar.ybp <= input$agerange2[2])

distbounds <- input$distmap_bounds

distlatRng <- range(distbounds$north, distbounds$south)

distlngRng <- range(distbounds$east, distbounds$west)

subset(dist2,

lat >= distlatRng[1] & lat <= distlatRng[2] &

lng >= distlngRng[1] & lng <= distlngRng[2]

)

})

140

# Remove extra .id column

rawdatamain <- reactive({

visibledatasets()[,-1]

})

rawdatadist <- reactive({

distributions()[,-1]

})

qsummary2 <- reactive({

distributions() %>%

group_by(Taxon) %>%

summarise("Min Lat" = format(min(lat), digits = 5, scientific=FALSE),

"Min Long" = format(min(lng), digits = 5, scientific=FALSE),

"Max Lat" = format(max(lat), digits = 5, scientific=FALSE),

"Max Long" = format(max(lng), digits = 5, scientific=FALSE),

"Record Count" = length(dataset),

"Unique Sites" = length(unique(site.name))

)

})

qsummary3 <- reactive({

141

visibledist() %>%

group_by(Taxon) %>%

summarise("Min Lat" = format(min(lat), digits = 5, scientific=FALSE),

"Min Long" = format(min(lng), digits = 5, scientific=FALSE),

"Max Lat" = format(max(lat), digits = 5, scientific=FALSE),

"Max Long" = format(max(lng), digits = 5, scientific=FALSE),

"Record Count" = length(dataset),

"Unique Sites" = length(unique(site.name))

)

})

output$quicksummary2 <- renderTable({

if (nrow(distributions()) == 0)

return(NULL)

qsummary2()

striped = TRUE,

rownames = FALSE,

spacing = "s",

align = "c"

)

output$quicksummary3 <- renderTable({

142

if (nrow(distributions()) == 0)

return(NULL)

qsummary3()

striped = TRUE,

rownames = FALSE,

spacing = "s",

align = "c"

)

domain <- levels(long_DF$Taxon)

pal <- colorFactor(c("#e6194B", "#3cb44b", "#ffe119", "#4363d8", "#f58231", "#42d4f4",

"#f032e6", "#fabebe", "#469990",

"#e6beff", "#9A6324", "#fffac8", "#800000", "#aaffc3", "#000075", "#a9a9a9",

"#000000"

domain = domain)

observe(

leafletProxy("distmap", data = distributions()

) %>%

clearMarkers() %>%

clearShapes() %>%

143

clearControls() %>%

addCircleMarkers(~lng, ~lat, radius = 2, color = ~pal(Taxon),

popup = paste(

distributions()$site.name, "
",

"Lat:", format(distributions()$lat, digits = 4, scientific=FALSE),

"Long:", format(distributions()$lng, digits = 4, scientific=FALSE), "
",

"Dataset ID:", distributions()$dataset, "
",

"Age (Calendar ybp):", format(distributions()$calendar.ybp, digits = 0, scientific =

FALSE),

label = NULL))%>%

addLegend("bottomright", pal = pal, values = ~Taxon,

title = "Currently Displayed Taxa",

labFormat = labelFormat(prefix = NULL),

opacity = 1

)

# Raw Data

# Main Tab

output$rawdata <- renderDataTable({

if (nrow(visibledatasets()) == 0)

return(NULL)

return(rawdatamain())

144

# striped = TRUE,

rownames = FALSE,

# spacing = "s",

# align = "c",

options = list(paging=FALSE, scrollX = TRUE)

)

# Distribution Tab

output$rawdata2 <- renderDataTable({

if (nrow(rawdatadist()) == 0)

return(NULL)

return(rawdatadist())

# striped = TRUE,

rownames = FALSE,

# spacing = "s",

# align = "c",

options = list(paging=FALSE, scrollX = TRUE)

)

# Clears all user selections and returns Main map to default observeEvent(

145

input$resetall, {

reset("site")

reset("regions")

reset("state")

reset("agerange")

reset("selectsite")

reset("confirmation")

output$mainmap <- renderLeaflet({

leaflet() %>%

clearMarkers() %>%

addProviderTiles(providers$CartoDB.Positron)%>%

setView(lng=-96, lat=41, zoom=3.5)

})

}

)

# Clears all user selections and returns Distribution map to default observeEvent(

input$resetall2, {

reset("taxa")

reset("agerange2")

output$distmap <- renderLeaflet({

leaflet() %>%

146

clearMarkers() %>%

addProviderTiles(providers$CartoDB.Positron)%>%

setView(lng=-96, lat=41, zoom=3.5)

})

}

)

output$d1 <- downloadHandler(

filename = function() {

paste("data-", Sys.Date(), ".csv", sep="")

content = function(file){

withProgress(message = 'Working on it', value = 0, {

vis <- get_dataset( as.numeric(visibledatasets()$dataset),

ageold = input$agerange2[2], ageyoung = input$agerange2[1] )

incProgress(1/2, detail = "Downloading")

down <- get_download(vis)

incProgress(2/2, detail = "Compiling")

visd <- compile_downloads(down)

147

write.csv(visd, file, row.names = FALSE)

})

}

)

output$d2 <- downloadHandler(

filename = function() {

paste("data-", Sys.Date(), ".csv", sep="")

content = function(file){

withProgress(message = 'Working on it', value = 0, {

dvis <- get_dataset( as.numeric(visibledist()$dataset),

ageold = input$agerange2[2], ageyoung = input$agerange2[1] )

incProgress(1/2, detail = "Downloading")

d.down <- get_download(dvis)

incProgress(2/2, detail = "Compiling")

visdata <- compile_downloads(d.down)

write.csv(visdata, file, row.names = FALSE)

})

148

}

)

output$d3 <- downloadHandler(

filename = function() {

paste("data-", Sys.Date(), ".csv", sep="")

content = function(file){

withProgress(message = 'Working on it', value = 0, {

dall <- get_dataset( as.numeric(distributions()$dataset),

ageold = input$agerange2[2], ageyoung = input$agerange2[1] )

incProgress(1/2, detail = "Downloading")

dall.down <- get_download(dall)

incProgress(2/2, detail = "Compiling")

alldata <- compile_downloads(dall.down)

write.csv(alldata, file, row.names = FALSE)

})

149

}

)

}

Get_downloads.R library(neotoma)

# Make a backup of the current data currentdata <- read.csv("nam whitmore full 22k.csv") write.csv(currentdata, paste("nam whitmore full 22k", Sys.Date(), ".csv", sep = ""), row.names =

FALSE) rm(currentdata)

# Get the full dataset from Neotoma alldata <- get_dataset(gpid = c("United States", "Canada"), datasettype = 'pollen', ageold =

22000)

alldownloads <- get_download(alldata)

alltaxa <- compile_taxa(alldownloads, "WhitmoreFull")

compiled <- compile_downloads(alltaxa)

write.csv(compiled, "nam whitmore full 22k.csv", row.names = FALSE)

150

Userguide.Rmd

--- title: "GeoPollen User Guide" author: "Hannah M Carroll" date: "2/24/2020" output:

html_document: default header-includes:

- \setlength{\parindent}{4em}

- \setlength{\parskip}{0em}

---

# GeoPollen Version 0.9

---

## Description

GeoPollen is a Shiny Dashboard application designed to allow for visualization and basic analysis of pollen datasets publicly available through the Neotoma Paleoecology Database

(https://www.neotomadb.org/) [(Williams et al.

2018)](https://www.cambridge.org/core/services/aop-cambridge- core/content/view/1E1C9EB07ADFF01182BCB69A08E1C755/S0033589417001053a.pdf/neoto ma_paleoecology_database_a_multiproxy_international_communitycurated_data_resource.pdf).

151

This project is not hosted by nor associated with Neotoma. GeoPollen is a standalone project and the application developers are solely responsible for its content.

## Authors

**Maintainer:** Hannah M. Carroll (https://orcid.org/0000-

0003-3343-3358)

Authors:

- Hannah M. Carroll

- Lynn G. Clark

- Alan D. Wanamaker (http://orcid.org/0000-0002-6560-6420)

## Bug Reports

Report bugs to https://github.com/hannahcarroll/GeoPollen/issues

Questions, suggestions, or feature requests should go directly to the developer and maintainer of

GeoPollen rather than GitHub:

152

---

## Overview

GeoPollen can be accessed via the web at https://geopollen.eeob.iastate.edu/ or downloaded from

GitHub and run locally via RStudio. Running

GeoPollen locally may be preferable if you intend to work with a large amount of data and/or use the application for complex processing, *i.e.*, performing analog calculations on more than several dozen datasets.

GeoPollen is arranged into tabs in a sidebar. The **Main** tab contains a map of all available datasets in the selected spatiotemporal range. The map defaults to being centered over the continental United States, and the time slider defaults to 10,000 Cal BP - 8,000 Cal BP. The map, time slider, and menus are *reactive*, meaning that as the user makes changes the application immediately updates itself based on the new input. The time slider can be operated by dragging one handle at a time to the selected range. If desired, the time slider can be dragged as a unit.

This makes it possible to visualize data in discrete time ranges across the full record.

drawing

153

The map is populated with individual *sites*, each of which returns one or more *datasets*,

*e.g.*, rows of pollen data. Each row represents a single time point, with metadata and pollen counts arranged in columns.

The datasets are assigned unique row names at app startup in order to make identifying and tracking samples easy. The row names are a concatenation of the site name, dataset ID, depth, and age (in Cal BP).

---

## Dataset Ages

Raw radiocarbon dates are updated to IntCal09 after [Reimer et al.

(2009)](https://journals.uair.arizona.edu/index.php/radiocarbon/article/view/3569). Calibrated ages reported by authors are not changed. Ages are stored in the raw data in a column named

"calendar.ybp." All original age information submitted by authors is retained in the raw data. All ages displayed in GeoPollen are in Cal BP.

---

## Tab Overview

154

Data can be selected in either the **Main** tab or the **Distribution** tab. The **Main** tab allows users to visualize the spatiotemporal distribution of available datasets and perform analog calculations at the community level. The **Distribution** tab allows users to visualize the spatiotemporal distribution of one or more individual taxa, rather than datasets or communities.

The maps are independent of each other and selections in one tab will not affect selections in the other.

---

### Main Tab

The **Main** tab allows users to view the spatiotemporal distribution of datasets available through the Neotoma Paleoecology Database. This is not meant to replicate the functionality of

Neotoma Explorer, Neotoma's native GUI, simply to provide a tool for analysis that obviates the need for users to write R code.

The **Main** tab provides basic controls for performing modern analog calculations. Clicking on a map marker will populate the drop down menu in the Analog Controls box with calendar ages of datasets available for that site in the user's specified age range. A scatterplot displays the site's age model (within the selected age range only). One or more datasets may be selected, although some training set diagnostics are only generated when three or more datasets are selected. The user's selected datasets become the training set for performing analog calculations.

155

drawing

When the Find Analogs button is clicked, the Squared Chord Distance is calculated between the training set and all other currently visible sites, also known at the test set. All diagnostics are returned to the **Analog Communities** tab.

drawing

If another site is selected after the Find Analogs button has been clicked, all previous selections are cleared.

Analog calculations are performed using package 'analogue' [(Simpson and Oksanen

2019,](https://cran.r-project.org/package=analogue) [Simpson

2007)](http://dx.doi.org/10.18637/jss.v022.i02). The current version is 0.17-3. For package documentation, see https://cran.r-project.org/web/packages/analogue/analogue.pdf

Squared Chord Distance is the distance metric most commonly used for pollen data. Package

'analogue' offers several options, but SCD is presently the only available metric in GeoPollen.

Future releases will expand this functionality. Squared Chord Distance values range from 0

(identical) to 2 (completely dissimilar).

A download button allows the user to download datasets currently in view. When a download button is clicked, a download progress bar appears in the bottom right corner. The download

156 handler uses package 'neotoma' [(Goring et al.

2015)](https://www.openquaternary.com/article/10.5334/oq.ab/) to send an API call to the

Neotoma Paleoecology Database. Datasets are downloaded and then compiled into a spreadsheet in .csv format. A dialogue box popup allows the user to select the destination folder.

drawing

### Analog Communities Tab

The **Analog Communities** tab is divided into three sub-tabs. The Minimum Dissimilarity sub-tab displays the user's training set metadata and diagnostic output from the *analog* function in package 'analogue'.

drawing

The Training Set Diagnostics sub-tab shows a curve of training set dissimilarities and an interactive Plotly heatmap based on the pairwise Squared Chord Distance. These tools enable the user to explore their selected training set and judge its suitability. Both plots may be saved from within GeoPollen.

drawing

157

In this example, the training set consists of Lima Bog, which is dataset 16129, ages 1668, 2166,

2368, and 2470 Cal BP.

The Analogs sub-tab displays an interactive heatmap comparing all training set samples to all test set samples, and a searchable table of the pairwise Squared Chord Distance between all selected sites. The user may download the heatmap in .png format directly from the app, or copy and paste the SCD matrix from the table. All raw data are reactively displayed in the Raw Data tab.

drawing

To inspect a portion of the heatmap more closely, simply click and drag a box over the desired area. Double click or press the Home button in the upper right corner of the figure to reset the view.

drawing

---

### Distribution Mapping Tab

The **Distribution Mapping** tab allows users to view the spatiotemporal distribution of one or more taxa and select which datasets to download. The left reactive table displays the full

158 spatiotemporal range of the selected taxa in the selected time period. The right reactive table displays spatiotemporal information for just those sites currently in view.

drawing

Two download options are available. Both use package 'neotoma' (Goring et al. 2015) to send an

API call to the Neotoma Paleoecology Database. The requested datasets are downloaded to a spreadsheet in .csv format and a dialogue box allows the user to choose where to save the file.

When a download button is clicked, a download progress bar appears in the bottom right corner.

---

### Raw Data Tab

The **Raw Data** tab is divided into two sub-tabs. The Main sub-tab reactively displays a searchable table of all data currently displayed on the **Main** tab. The Distribution Mapping sub-tab reactively displays a searchable table of all data currently displayed on the

**Distribution Mapping** tab. Data in the Main tab are displayed in a wide format, while data in the Distribution Mapping tab are displayed in long format. The sub-tabs are independent of each other and choices made in one tab do not affect the data displayed in the other.

---

159

### Citations Tab

The **Citations** tab displays general information for citing the Neotoma Paleoecology

Database, constituent databases, and dataset authors. This is periodically updated, but for the most current information, always refer to https://www.neotomadb.org/data/category/use

ALL USES OF DATA OBTAINED THROUGH GEOPOLLEN MUST ADHERE TO THE

NEOTOMA DATA USE AND EMBARGO POLICY.

---

### References

Goring, S., Dawson, A., Simpson, G. L., Ram, K., Graham, R. W., Grimm, E. C., & Williams, J.

W.. (2015). neotoma: A

Programmatic Interface to the Neotoma Paleoecological Database, 1(1), Art. 2. DOI: http://doi.org/10.5334/oq.ab

Reimer, P.J., Baillie, M.G., Bard, E., Bayliss, A., Beck, J.W., Blackwell, P.G., Ramsey, C.B.,

Buck, C.E., Burr, G.S., Edwards, R.L. and Friedrich, M. (2009). IntCal09 and Marine09 radiocarbon age calibration curves, 0-50,000 years cal BP. Radiocarbon, 51(4), 1111--1150.

160

Simpson, G.L. and Oksanen, J. (2019). analogue: Analogue matching and Modern Analogue

Technique transfer function models. (R package version 0.17-3 ). (https://cran.r-project.org/package=analogue).

Simpson, G.L. (2007). Analogue Methods in Palaeoecology: Using the analogue Package.

Journal of Statistical Software,

22(2), 1--29.

Williams, J.W., Grimm, E.G., Blois, J., Charles, D.F., Davis, E., Goring, S.J., Graham, R.,

Smith, A.J., Anderson, M., Arroyo-Cabrales, J., Ashworth, A.C., Betancourt, J.L., Bills, B.W.,

Booth, R.K., Buckland, P., Curry, B., Giesecke, T., Hausmann, S., Jackson, S.T., Latorre, C.,

Nichols, J., Purdum, T., Roth, R.E., Stryker, M., Takahara, H., (2018). The Neotoma

Paleoecology Database: A multi-proxy, international community-curated data resource.

Quaternary Research 89, 156--177.

Citeneotoma.Rmd

--- title: author: date: "Updated 2/21/2020" output: html_document

---

# How to Cite the Neotoma Paleoecology Database

161

### All uses of data from the Neotoma Paleoecology Database should be accompanied by an acknowledgement of the Neotoma community and contributors, as well as the individual dataset contributors

Click here for more information: https://www.neotomadb.org/data/category/use

To cite the Neotoma Paleoecology Database, use the preferred citation format:

Williams, J.W., Grimm, E.G., Blois, J., Charles, D.F., Davis, E., Goring, S.J., Graham, R.,

Smith, A.J., Anderson, M., Arroyo-Cabrales, J., Ashworth, A.C., Betancourt, J.L., Bills, B.W.,

Booth, R.K., Buckland, P., Curry, B., Giesecke, T., Hausmann, S., Jackson, S.T., Latorre, C.,

Nichols, J., Purdum, T., Roth, R.E., Stryker, M., Takahara, H., 2018. The Neotoma Paleoecology

Database: A multi-proxy, international community-curated data resource. Quaternary Research

89, 156-177.

To cite this dashboard, please use the following:

---

#### GeoPollen was developed in R version 3.6.2 and RStudio version 1.2.5033

**Analog calculations are performed using package 'analogue':**

162

Simpson, G.L. and Oksanen, J. (2019). analogue: Analogue matching and Modern Analogue

Technique transfer function models. (R package version 0.17-3). (https://cran.r- project.org/package=analogue).

Simpson, G.L. (2007). Analogue Methods in Palaeoecology: Using the analogue Package

Journal of Statistical Software, 22(2), 1--29.

**Downloads are handled by package 'neotoma':**

Goring, S., Dawson, A., Simpson, G. L., Ram, K., Graham, R. W., Grimm, E. C., & Williams,

J. W.. (2015). neotoma: A Programmatic Interface to the Neotoma Paleoecological Database,

1(1), Art. 2. DOI: http://doi.org/10.5334/oq.ab

---

#End

163

APPENDIX B. PHYTOLITHS OF THE FLORA OF MISSOURI: A REFERENCE COLLECTION

A reference collection of phytoliths from representatives of all silica phytolith-bearing families of Missouri was produced in order to support the phytolith analysis reported in Chapter

4 of this dissertation. Selection of Missouri taxa followed Steyermark’s Flora of Missouri

(Yatskievych and Steyermark 1999), which was then narrowed to those taxa listed by Piperno

(2006) as being known, likely, or potential phytolith producers. Only those which have been confirmed not to produce phytoliths were excluded from sampling.

Specimens were obtained from the Ada Hayden Herbarium (ISC) at Iowa State

University with the permission of Deborah Lewis, Curator and Dr. Lynn Clark, Director.

Sampling and processing of the material was completed with the help of Kathryn Holmes.

Phytolith extraction from herbarium material was performed using the wet oxidation method of

Piperno (2006). Extracted material was mounted in Permount on glass slides. A copy of the reference collection is housed in the Ada Hayden Herbarium. Extra phytolith material is stored dry in borosilicate glass vials.

Images of reference slides were taken in the laboratory of Dr. Beth Caissie in the

Geological and Atmospheric Sciences Department at Iowa State University. Images were taken with the assistance of Jacqueline Galang. Images were taken on a Leica upright microscope with a microscope-mounted camera. Image magnification varies between 400x and 1000x; those at

600x and 1000x are taken with oil immersion. All images are taken with differential interference contrast (DIC). Considerable effort was made to capture all distinct phytolith morphotypes produced by each taxon.

Not all taxa which were sampled produced usable phytoliths. The tables that follow give those taxa which did produce usable phytoliths. For grasses, see Table 1; for non-grass

164 monocots, see Table 2; for conifers, see Table 3; for pteridophytes, see Table 4; for non-aster dicots, see Table 5; for asters, see Table 6. Images for each taxon have been submitted to the

PhytCore Database (Albert et al. 2016).

Table B-1: Accession numbers and taxonomic information for grasses included in the phytolith reference collection.

Grasses Family Subfamily Tribe Taxon Accession Number Poaceae Aristidoideae Aristideae Aristida adscensionis ISC-281706 Poaceae Bambusoideae Arundinarieae Arundinaria gigantea ISC-412378 Poaceae Arundinoideae Molinieae Phragmites australis ISC-173379 Poaceae Chloridoideae Aeluropodeae Distichlis spicata ISC-415743 Poaceae Chloridoideae Cynodonteae Bouteloua curtipendula ISC-434241 Poaceae Chloridoideae Cynodonteae Buchloe dactyloides ISC-447507 Poaceae Chloridoideae Eragrostideae Muhlenbergia capillaris ISC-167339 Poaceae Chloridoideae Zoysieae Spartina pectinata ISC-282585 Poaceae Chloridoideae Zoysieae Sporobolus heterolepis ISC-362129 Poaceae Danthonioideae Danthonieae Danthonia spicata ISC-266809 Poaceae Oryzoideae Oryzeae Leersia oryzoides ISC-221986 Poaceae Oryzoideae Oryzeae Zizania palustris ISC-192243 Poaceae Panicoideae Andropogoneae Andropogon gerardii ISC-221746 Poaceae Panicoideae Andropogoneae Bothriochloa saccharoides ISC-270195 Poaceae Panicoideae Andropogoneae Schizachyrium scoparium ISC-221301 Poaceae Panicoideae Chasmanthieae Chasmanthium latifolium ISC-265983 Poaceae Panicoideae Paniceae Cenchrus longispinus ISC-229789 Poaceae Panicoideae Paniceae Echinochloa crusgalli ISC-255777 Poaceae Panicoideae Paniceae Setaria parviflora ISC-244830 Poaceae Panicoideae Paspaleae Paspalum bifidum ISC-213589 Poaceae Pooideae Aveneae Holcus lanatus ISC-285122 Poaceae Pooideae Brachyelytreae Brachyelytrum erectum ISC-180038 Poaceae Pooideae Bromeae Bromus latiglumis ISC-424496 Poaceae Pooideae Diarrheneae Diarrhena obovata ISC-363517 Poaceae Pooideae Meliceae Melica nitens ISC-228635 Poaceae Pooideae Poeae Alopecurus aequalis ISC-425473 Poaceae Pooideae Poeae Calamagrostis canadensis ISC-282143 Poaceae Pooideae Poeae Calamovilfa longifolia ISC-151642 Poaceae Pooideae Poeae Cinna arundinacea ISC-426325 Poaceae Pooideae Stipeae Piptatherum racemosum ISC-222228 Poaceae Pooideae Stipeae Hesperostipa spartea ISC-221870 Poaceae Pooideae Triticeae Elymus canadensis ISC-175804 Poaceae Pooideae Triticeae Hordeum jubatum ISC-408977

165

Table B-2: Accession numbers and taxonomic information for non-grass monocotyledons included in the phytolith reference collection.

Non-Grass Monocots Family Taxon Accession Number Commelinaceae Commelina virginica ISC-213362 Cyperaceae Bolboschoenus fluviatilis ISC-445691 Cyperaceae Carex missouriensis IA-91099 Cyperaceae Dulichium arundinaceum ISC-440058 Cyperaceae Eleocharis acicularis ISC-440476 Cyperaceae Fimbristylis vahlii ISC-377792 Cyperaceae Scirpus atrovirens ISC-409265 Juncaceae Juncus acuminatus ISC-211650 Juncaceae Luzula campestris ISC-154446 Marantaceae Thalia dealbata ISC-249557 Orchidaceae Calopogon tuberosus ISC-249901 Orchidaceae Liparis liliifolia ISC-325934

Table B-3: Accession number and taxonomic information for the conifer included in the phytolith reference collection.

Conifers Family Taxon Accession Number Cupressaceae Juniperus virginiana ISC-331644

Table B-4: Accession numbers and taxonomic information for the pteridophytes included in the phytolith reference collection.

Pteridophytes Family Taxon Accession Number Equisetaceae Equisetum arvense ISC-421039 Selaginellaceae Selaginella rupestris ISC-424285

Table B-5: Accession numbers and taxonomic information for non-aster dicotyledons included in the phytolith reference collection.

Non-aster dicots Family Taxon Accession Number Acanthaceae Justicia ovata ISC-329330 Aceraceae Acer negundo ISC-416852 Boraginaceae Cynoglossum virginianum ISC-261137 Boraginaceae Hackelia virginiana ISC-207355 Boraginaceae Mertensia virginica ISC-445839 Boraginaceae Myosotis verna ISC-332033 Boraginaceae Onosmodium molle ISC-5128 Caprifoliaceae Symphoricarpos occidentalis ISC-397292

166

Table B-5 (continued)

Family Taxon Accession Number Caprifoliaceae Triosteum perfoliatum ISC-345700 Cucurbitaceae Cucurbita foetidissima ISC-103377 Cucurbitaceae Melothria pendula ISC-230346 Cucurbitaceae Sicyos angulatus ISC-181586 Fagaceae Fagus grandifolia ISC-162825 Fagaceae Quercus alba ISC-151411 Magnoliaceae Magnolia acuminata ISC-74705 Moraceae Morus rubra ISC-74346

Ulmaceae Celtis occidentalis ISC-97693

Table B-6: Accession numbers and taxonomic information for asters included in the phytolith reference collection.

Asters Family Subfamily Tribe Taxon Accession Number Asteraceae Asteroideae Anthemideae Achillea millefolium ISC-420366 Asteraceae Asteroideae Anthemideae Artemisia campestris ISC-452475 Asteraceae Asteroideae Astereae Erigeron annuus ISC-207068 Asteraceae Asteroideae Eupatorieae Brickellia eupatorioides ISC-249630 Asteraceae Asteroideae Eupatorieae Liatris mucronata ISC-206084

References

Albert, R. M., J. A. Ruíz, and A. Sans. 2016. PhytCore ODB: A new tool to improve efficiency in the management and exchange of information on phytoliths. Journal of Archaeological Science 68:98–105.

Piperno, D. R. 2006. Phytoliths: a comprehensive guide for archaeologists and paleoecologists. Rowman Altamira.

Yatskievych, G. A., and J. A. Steyermark. 1999. Steyermark’s flora of Missouri. Missouri Department of Conservation.

167

APPENDIX C. USING LIGHT STABLE ISOTOPES TO ASSESS STREAM FOOD WEB ECOLOGY IN A GENERAL ECOLOGY LABORATORY COURSE

Hannah M. Carrolla, Derek D. Houstonb, Suzanne Ankerstjernec, and Alan D. Wanamaker, Jr.c aDepartment of Ecology, Evolution, and Organismal Biology, Iowa State University, Ames, IA,

USA; bDepartment of Natural and Environmental Sciences, Western Colorado University,

Gunnison, CO, USA; cDepartment of Geological and Atmospheric Sciences, Iowa State

University, Ames, IA, USA

Modified from a manuscript published in Journal of Biological Education

Abstract

Stable isotopes in natural materials provide a powerful way to study energy flow in many systems and are widely used in fields such as archaeology, ecology, forensics, geochemistry, geology, oceanography, paleoecology and paleoclimatology. Based on the manner in which stable isotopes fractionate in natural systems, they allow scientists to address a wide array of research topics ranging from tracking climatic shifts, ascertaining organisms’ migratory patterns, matching organisms to their diets and/or environments, assessing food web bioenergetics, documenting ecosystem changes through time, measuring soil carbon budgets and soil microbial activity, etc. Hence, it may be a viable option to use stable isotopes to investigate stream and food web ecology with students. Students with no prior experience working with stable isotopes successfully met learning objectives by completing the requisite field and laboratory protocols, analyzing data, interpreting results, and communicating their findings in a report modelled after a peer-reviewed scientific journal article. While this activity focused on food web ecology in a stream ecosystem, the method is repeatable, cost-effective, and can be modified relatively easily to evaluate food webs in virtually any other ecosystem.

168

Introduction

Stable isotopes are alternate forms of atoms that differ in the number of neutrons contained within their nuclei. Stable isotopes maintain the same chemical properties of their elements, but differ in their atomic mass (Fry 2006). They are non-radioactive atoms that do not experience radioactive decay (as opposed to radioactive isotopes, such as carbon-14, that are commonly used in radiometric dating). Stable isotopes are powerful tools used in many scientific disciplines including community ecology, climate science, fisheries biology, paleoecology, geology, forensic science, archaeology, soil science, and many others. Fractionation processes partition heavy and light isotopes in natural materials, largely due to mass effects (Sharp 2007), therefore the stable isotope signatures in such materials provide clues about energy flow and/or environmental conditions. Thus, stable isotopes are commonly used to track changes in climate

(Cerling 1984; Lipp et al. 1991; McDermott 2004; West et al. 2006; Baker et al. 2017), ascertain migratory patterns (Rubenstein et al. 2002; Rubenstein and Hobson 2004; Hobson 2016), match organisms to their environments (Bearhop et al. 2004; Hogan et al. 2014) or diets (Hilderbrand et al. 1996; Richards et al. 2000; Bearhop et al. 2003; Divine et al. 2017; Patterson et al. 2019), assess food web bioenergetics (Bunn and Boon 1993; Hobson, Piattt, and Pitocchelli 1994;

Vander Zanden, Casselman, and Rasmussen 1999; Hershey et al. 2017), document ecosystem and environmental changes through time (O’Reilly et al. 2003; Reynolds et al. 2016, 2017;

Whitney et al. 2019), measure soil carbon budgets (Hsieh and Yapp 1999; Bridgham et al. 2006), assess soil microbial activity (Boschker and Middelburg 2002; Steinbeiss, Gleixner, and

Antonietti 2009), and other topics.

Commonly used light stable isotopes include hydrogen, carbon, oxygen, nitrogen, and sulfur. Stable isotopes are measured as a function of the ratio of the more common isotope to the less common isotope, relative to the same ratio measured in an international standard. This value

169 is multiplied by 1,000, and the results are reported in parts per thousand (permil; ‰).

Researchers can examine isotopic ratios, designated as delta values (δ) (see Equation 1),

Eq. 1:  = ((푅푥 − 푅푠푡푑)/푅푠푡푑) ∗ 1000

where R is the ratio of the abundance of the heavy to light isotope (e.g. 2H/1H, 13C/12C,

18O/16O, 15N/14N, 34S/32S), x denotes sample, and std is the abbreviation for standard, to answer questions of interest (Peterson and Fry 1987; Fry 2006). For example, the δ18O value in mollusk shell material is commonly used to constrain seawater temperature conditions at the time of deposition when the isotopic composition of the water is known or can be reliably estimated

(Epstein et al. 1953; Wanamaker et al. 2007). Another example is the use of δ13C and δ15N values to infer animal diets and evaluate ecosystem trophic structure (Hershey et al. 2017).

Using carbon from a plant as an example, researchers would place a sample in an elemental analyzer along with an international standard for carbon, on the Vienna Pee Dee

Belemnite (VPDB) scale. The elemental analyzer combusts the sample producing CO2, and a

13 16 mass spectrometer measures the relative intensities of the rare isotope ( C O2; mass 45) and the

12 16 common isotope ( C O2; mass 44) via Faraday cups, and provides an isotopic ratio based on the relative intensities of mass 45 and mass 44 derived from the plant’s tissues. The instruments then repeat the measurements on a number of international isotopic standards with well-constrained

δ13C values and are used to place the samples on the international isotope scale, VPDB. Organic carbon is almost always isotopically negative. For example, the tissues of plants range from a

13 δ C of −10‰ to −29‰ (O’Leary 1988), depending on the kind of plant (C3, C4, or CAM) and its environment. This is because plants preferentially incorporate the lighter isotope of carbon, which is the most common in nature (more than 98% of the global carbon pool). The negative value simply shows that there is less of the heavier (rare) isotope relative to the lighter (common)

170 one in the plant’s tissues when compared to the standard. This property whereby one isotope of an element is incorporated relatively more than another is called fractionation, and can be utilized to understand the flow of energy through systems, including food web ecology.

Generally speaking, lighter isotopes have weaker bonds than heavier isotopes (Sharp 2007), thus it is thermodynamically easier for 12C rather than 13C to be incorporated into the leaf of a plant during photosynthesis.

Organisms within an ecosystem are often grouped into trophic levels that are descriptive of how they attain their energy. Producers (autotrophs) comprise the lowest trophic level of a system and convert light (or chemical) energy into usable forms of chemical energy (e.g. sugars).

Consumers (heterotrophs) are organisms that gain energy by consuming producers, and can be classified as primary consumers (herbivores, planktivores), secondary consumers (carnivores), tertiary consumers (sometimes carnivores, sometimes omnivores), and apex predators that feed on primary, secondary, and tertiary consumers if there is enough energy in the system to support that many trophic levels. In most cases, approximately 10% of the energy at one trophic level is incorporated into biomass at the next highest trophic level, although that varies from system to system (Molles and Sher 2019). The isotopic signatures of primary producers (e.g. algae) are determined by dissolved inorganic carbon (DIC), and thus it is ultimately the system’s DIC signature which is reflected in the isotopic signatures of all producers and consumers within the food web (Finlay 2001).

Freshwater ecosystems receive organic matter from terrestrial environments that is incorporated as allochthonous material (e.g. dead leaves, fallen branches), otherwise described as coarse particulate organic matter (CPOM), and impacts energy flow (Vannote et al. 1980; Junk,

Bayley, and Sparks 1989). Organisms called shredders (e.g. crane flies, some caddisflies, some

171 stoneflies, some midges) consume CPOM (Cummins and Klug 1979; Cummins et al. 1989), so there is a large detritus-based component of many freshwater stream food webs. Shredders convert CPOM into fine particulate organic matter (FPOM) that may be consumed by downstream collectors (e.g. some caddisflies, some beetles, some dipterans, some midges)

(Vannote et al. 1980). Physical breakdown of CPOM and leaching of some solutes can also result in dissolved organic matter (DOM) that can be consumed by zooplankton (e.g. diatoms, water fleas). Freshwater ecosystems also contain autochthonous materials (e.g. microbes, plankton, algae, aquatic plants) that originate within the system. Scrapers (e.g. some snails, some caddisflies, some fish) and grazers (e.g. mayflies, some beetles, some snails, suckers) often consume these materials (Vannote et al. 1980; Junk, Bayley, and Sparks 1989; Thorp and Delong

1994). Thorp and Delong (1994) postulated that autochthonous primary production contributed a higher proportion of carbon to consumers than had been previously recognized. Of course, there are secondary consumers, tertiary consumers and apex predators within these systems as well

(e.g. dragonflies, some midges, insectivorous fishes, piscivorous fishes, birds of prey [such as eagles, cranes, and osprey], and some mammals [such as raccoons, river otters, bears, and humans]). A hypothetical stream food web is diagrammed in Figure C-1.

Regardless of whether an ecosystem is terrestrial or aquatic, the flow of carbon between trophic levels produces isotopic enrichment. Biological fractionation occurs initially when plants photosynthesize and incorporate the products of photosynthesis into their tissues. Plant tissues are eaten by grazers, which in turn preferentially incorporate the lighter isotopes from the plant tissues into their own tissues. This continues up through trophic levels in a predictable sequence of enrichment of about 1‰ for δ13C and 3‰ for δ15N at each step (DeNiro and Epstein 1976), and means that the trophic position and diet of an organism is reflected in the isotopic

172 composition of its tissues. However, the absolute δ13C signature is heavily influenced by longitudinal gradients, seasonality, and geography (Kobayashi et al. 2011).

Figure C-1: A graphical representation of a generic aquatic food web depicting energy flow across multiple trophic levels (modified after Merritt and Cummins 1996). Organisms are not drawn to scale. Abbreviations are as follows: CPOM – coarse particulate organic matter; DOM – dissolved organic matter; FPOM – fine particulate organic matter.

In this exercise, we engaged biology undergraduate students in an active learning experience using stable isotopes to evaluate river food web ecology in a Colorado (CO) tailwater fishery. Stable isotope analysis was chosen as an experiential education tool because it is a powerful, yet simple and relatively inexpensive technique that is easily understood by undergraduate students. Food web ecology is a fundamental concept in community ecology, and streams are convenient systems in which to demonstrate connectivity among trophic levels because they tend to be more contained than terrestrial ecosystems. Moreover, we have found that student interest and engagement in aquatic ecology is typically high, particularly in systems with conservation and/ or management implications. The activity centers on the premise that

173

δ13C and δ15N can be quantified and are useful in evaluating organismal trophic positions. The approach was successfully implemented in an upper division ecology laboratory course at

Western Colorado University, Gunnison, CO, USA, and trained students to be proficient in the use of stable isotopes, increased their understanding of aquatic ecology, enhanced their data management and analytical skills, and refined their scientific communication skills. While this activity was conducted in the southern Rocky Mountains of Colorado, it can easily be adapted to other ecosystems proximate to other institutions.

Student Learning Outcomes (SLOs)

Students will:

(1) demonstrate proficiency in sampling aquatic organisms using different techniques

including drift nets, kick nets, angling, and plankton tows (optional).

(2) classify aquatic organisms to biological order (e.g. Amphipoda, Annelida,

Ephemeroptera, Plecoptera), family (e.g. Chironomidae), or species (e.g. Salmo

trutta).

(3) designate the trophic level of sampled aquatic organisms (e.g. producers, primary

consumers, secondary consumers, tertiary consumers, apex predators, detritus).

(4) demonstrate safe laboratory practices (wearing appropriate personal protective

equipment while washing samples under a fume hood), and avoiding cross-

contamination of samples while preparing them for submission.

(5) analyze data in the open-source statistical software program R, reassess a priori

assumptions, and draw conclusions based on the data.

(6) communicate their findings in a scientific report modelled after a peer-reviewed

journal article that includes background literature, a description of the methods

174

employed, the outcomes of statistical analyses, graphical representations of the data,

discussion/conclusions, and a literature cited section.

Optional (could be assessed via a pre-laboratory worksheet, but were not when this exercise was implemented):

(7) define common terminology used in stream ecology: allochthonous/autochthonous

material; coarse particulate organic matter (CPOM)/fine particulate organic matter

(FPOM)/dissolved organic matter (DOM); shredders/collectors/grazers/scrapers.

(8) define common terminology used in stable isotopes analyses: delta values, permil,

fractionation, elemental analyzer, mass spectrometry.

In this laboratory exercise, we reconfirmed the applicability of stable isotopes to stream food web ecology as a learning tool. Our approach was as follows: (1) sample aquatic organisms using a variety of techniques, (2) use safe laboratory techniques to wash and prepare samples for analysis, (3) analyze stable isotope data using open-source statistical software, (4) reassess a priori trophic level designations and draw conclusions.

Procedures

Site Selection

Aquatic organisms were sampled from the catch-and-release area on the Taylor River

(Figure C-2), just below Taylor Park Reservoir, approximately 22 km northeast of Almont, CO,

USA. This site was chosen because (1) it is a tailwater fishery with relatively constant seasonal water temperatures and is accessible year-round (we conducted sampling in February when most other water bodies in the area were frozen over), (2) it is not a highly productive ecosystem, so is relatively depauperate and capturing a high proportion of its biodiversity was likely, (3) it is in close proximity (~50 km) to Western Colorado University campus where students engaging in

175 the exercise were enrolled, (4) it is a popular sport fishery that many Western Colorado

University students are familiar with and excited to learn more about.

Field Sampling

Because we sought to collect organisms from several trophic levels, we employed an array of sampling methods (SLO1). These are outlined as follows (see also Supplemental File 1):

Angling

A self-selected group of five students with their own fishing equipment and valid

Colorado fishing licenses attempted to catch trout (the presumed apex predator of the system) using artificial flies and lures. If successful, they were instructed to take a small fin clip

(approximately 1 cm x 0.5 cm), which grows back, then immediately release the fish back into the river. Students were instructed to use forceps to handle fin clips so as not to risk contaminating the sample with oils from their fingers, then to place the fin clip into an appropriately labelled plastic bag. While it has been demonstrated that different fish tissues (e.g. muscle, fin, liver, gonad) exhibit variability in isotopic signatures (Jardine et al. 2005), we opted for the least intrusive, non-lethal method for acquiring tissue.

176

Figure C-2: Students sampling aquatic organisms from the catch-and-release section of the Taylor River, CO. Top left: Students receiving instruction (from DDH) on how to deploy drift nets. Top right and bottom left: Groups of students picking through kick net samples and sorting organisms by taxonomic group and/or presumed trophic level. Bottom right: Students pulling drift nets to conclude the sampling exercise.

Macroinvertebrate Sampling

The remaining students worked in teams to set three drift nets downstream of the sampling area, where they were left to collect materials carried by the current for the duration of the sampling activity (~1 hour). Once the drift nets were deployed, pairs of students sampled aquatic macroinvertebrates using kick nets. The technique for sampling with kick nets is fairly simple: students were instructed to place the net perpendicular to the flow of the river, then kick along the bottom of the river so that rocks overturned and the current carried any dislodged macroinvertebrates into the net. This activity was conducted for 30 seconds per sampling unit, after which the sample was transferred from the kick net to a sorting tray. Macroinvertebrates,

177 algae and detritus were sorted according to taxonomic group (e.g. Plecoptera, Ephemeroptera,

Chironomidae; SLO2), then removed using forceps and placed into individually labelled plastic bags. Each pair of students repeated these steps three times, but organisms were pooled (within student working groups) by their taxonomy (i.e. each group pooled all their stoneflies into one bag, all their mayflies into another, algae into a third, detritus into a fourth, and so on). When kick net sampling was completed, drift nets were checked and samples were collected from them and sorted in the same manner. Samples collected from kick nets and drift nets were kept separate. All plastic bags containing biological samples were packed in snow to keep samples cold during transport back to the laboratory. Upon return to the institution, samples were frozen at −20°C for one week, after which they were thawed and further sorted (see Laboratory

Techniques below). Cold storage of samples is only necessary if they will not be processed within the same laboratory period.

Plankton Sampling (Optional)

The sun set and temperatures dropped too quickly to sample using plankton tows while in the field in February, but this method could easily be added to the exercise at different times of year, or even by assigning a subset of students to conduct plankton tows instead of sampling with kick nets. Groups of 2–3 students could conduct oblique plankton tows wherein a plankton tow net is pulled through the water at the same depth for a known distance (see Supplemental File 1).

This allows students to calculate the total volume of the water sampled. Once the tow is complete, students use a wash bottle to wash any plankton that is clinging to the sides of the net into the collection jar. These samples need to be sorted in the laboratory under a dissecting microscope.

178

Laboratory Techniques

Sorting

Samples were already sorted by trophic level as best as students could ascertain in the field (see Macroinvertebrate Sampling above). However, it was critically important for the success of the exercise for them to refine sample identifications to ensure that samples were isolated from those of other trophic levels. This required students to pick through samples under dissecting microscopes and make accurate identifications (SLO2) and preliminary trophic level designations (SLO3). Prior to sorting, students were instructed to wear gloves when handling specimens and reminded of the importance of not touching the samples bare-handed so as not to contaminate them with oils from their skin.

Students used plastic forceps and weigh boats (again, taking care not to touch either barehanded) to sort samples into the following categories: detritus (sticks, leaves, etc.), algae, midges, mayflies, stoneflies, amphipod crustaceans, fish fin clips. Each sample was then assigned a unique identification code that included a group number, a tentative trophic level, and a sample number (SLO4; Table C-1). Weigh boats used in sorting, and glass Petri dishes used for washing (see below) were labelled accordingly using a permanent marker.

Prior conversations with an aquatic ecologist familiar with the system (Dr. Kevin

Alexander, Western Colorado University, personal communication with DDH) affirmed that the sampling area was relatively depauperate. Therefore, taxonomic identifications were able to be very broad. For example, stoneflies can be categorized as predators, shredders, collectors, and grazers in various areas, depending on the species that occupy those areas, but in this system all stoneflies were likely to be grazers.

179

Table C-1: Biological samples students acquired from the catch-and-release area of the Taylor River, below Taylor Park Reservoir, CO. Samples were assigned a taxonomic classification, categorized by trophic level, and assigned a unique sample identification number.

Sample Group Presumed Trophic Biological Sample ID Common # # Level (by Students) Classification Name 1 1 Secondary Consumer Annelida G1PAANS1 Leech 2 1 Grazer Amphipoda G1GRAMS2 Scud1 3 1 Grazer Plecoptera G1GRPLS3 Stonefly 4 1 Grazer Ephemeroptera G1GREPS4 Mayfly 5 1 Collector Chironomidae G1COCHS5 Midge2 6 1 Detritus Unidentified Plant G1DEDES6 Twig 7 2 Detritus Unidentified Plant G2DEDES1 Twig 8 2 Grazer Chironomidae G2GRCHS2 Midge 9 2 Grazer Amphipoda G2GRAMS3 Scud1 10 2 Grazer Ephemeroptera G2GREPS4 Mayfly 11 2 Grazer Plecoptera G2GRPLS5 Stonefly 12 2 Producer Algae G2PRALS6 Algae3 13 3 Shredder Plecoptera G3SHPLS1 Stonefly 14 3 Collector Chironomidae G3COCHS2 Midge4 15 3 Shredder Ephemeroptera G3SHEPS3 Mayfly 16 3 Producer Algae G3PRALS4 Algae 17 3 Detritivore Amphipoda G3DEAMS5 Scud1 18 3 Apex Predator Salmonidae G3APSAS6 Brown Trout 19 4 Producer Algae G4PRALS1 Algae3 20 4 Grazer Ephemeroptera G4GREPS2 Mayfly5 21 4 Detritus Unidentified Plant G4DEDES3 Twig6 22 4 Grazer Chironomidae G4GRCHS4 Midge4 23 5 Detritus Unidentified Plant G5DEDES1 Twig6 24 5 Producer Algae G5PRALS2 Algae 25 5 Shredder Trichoptera G5SHTRS3 Caddisfly 26 5 Grazer Chironomidae G5GRCHS4 Midge5 1 Amphipod samples that were pooled by the instructor after drying to ensure that the minimum quantity requirement was met 2 Sample included a single midge that did not meet the minimum size requirements, but also appeared to be a different species from other samples, so this sample was omitted from further analysis 3 Algal samples that were pooled by the instructor after drying to ensure that the minimum quantity requirement was met 4 Midge samples that were pooled by the instructor after drying to ensure that the minimum quantity requirement was met 5 Misidentified samples – these samples were the same species of stonefly, not midges or mayflies, and were pooled by the instructor after drying 6 Detritus samples that were pooled by the instructor after drying to ensure that the minimum quantity requirement was met

180

Washing

Samples were placed into labelled glass Petri dishes according to biological classification/presumed trophic level. Students donned gloves and safety goggles, then transferred Petri dishes containing samples to a fume hood where they were repeatedly washed with a 2:1 chloroform:methanol solution, taking care to keep track of sample IDs during the process (SLO4). Instructors should remind students that the wash solution will remove permanent marker, so they must take care not to wash away labels.

The washing procedure was as follows: (1) Students transferred 10 mL of the 2:1 chloroform:methanol solution into a labelled Petri dish containing a sample using a glass transfer pipette. (2) Using metal tweezers that had been pre-washed with the solution (the wash solution reacts with plastic tools, hence the need to use metal and glass, respectively), students gently agitated the samples for 30 seconds to remove lipids and other surface contaminants (e.g. dirt, pollen) according to previously published methods (Wassenaar and Hobson 1998). (3) Students then transferred the dirty wash solution to a glass waste container in the fume hood using the same glass transfer pipette, and then discarded the dirty pipette into an appropriately labelled broken glass waste container. (4) Using clean glass transfer pipettes for each subsequent wash, students washed the samples twice more, agitating the sample for 30 seconds each time. Please note that if conducting this exercise in a region in which carbonate bedrock is present (or suspected to be present), benthic algae samples must be washed in 10% hydrochloric acid to remove carbonate contamination, then rinsed with distilled water several times. This step was unnecessary in our exercise because the Taylor River bedrock is granitic.

Drying

All samples were left to dry in the Petri dishes under the fume hood (a dust free environment) for three days. Once dry, the instructor transferred the samples to a low-

181 temperature drying oven, where they were dried for 48 hours at 50-60°C (anywhere from between 24-48 hours should suffice). Dried samples were then transferred by the instructor to 1- dram glass vials, sealed, labelled, and shipped to the Stable Isotope Paleo Environments

Research Group (SIPERG) Laboratory at Iowa State University (ISU), Ames, Iowa, USA, for further processing. Some samples were pooled at the instructor’s discretion, after checking confirming students’ taxonomic identifications, to ensure that adequate amounts of samples (0.5-

1.0 g of dried tissue) were submitted as well as to reduce costs (see Table C-1).

Iowa State University SIPERG Stable Isotope Laboratory

Once received, samples were placed into tin capsules and immediately sealed. Carbon

(δ13C) and nitrogen (δ15N) stable isotope analyses were conducted on a Costech Elemental

Analyzer attached to a Finnigan Delta Plus XL mass spectrometer in continuous flow mode. δ13C was corrected according to the international VPDB standard, and δ15N was corrected via isotopic reference materials (Air). Corrections were made using a regression method, and results reported in permil (‰). Percent concentration (%) was calculated using the peak intensity of the sample against well-characterized (C:N) acetanilide standards. Analytical uncertainty at 1σ was

±0.11‰for C and ±0.09‰for N.

Instructions for sorting and processing samples in the laboratory were provided to students in a handout (see Supplemental File 2).

Data Analyses

Statistical analyses and graphing were conducted in R v3.2.2 (R Core Team 2015) using the packages ggplot2 v.2.1.0 (Wickham et al. 2009) and mgcv v.1.8 (Wood 2006). Students completed an R tutorial covering basic statistics earlier in the semester. Students were assigned to conduct numerous tasks in an R notebook tutorial (Supplemental File 3) including loading the data, creating a dual isotopes plot, plotting histograms, assessing distributions, removing outliers,

182 plotting a pairwise Euclidean distance matrix, conducting cluster analyses, and interpreting figures (SLO5). Pairs of students worked through the tutorial together, and turned in completed versions with properly annotated answers to questions that were then graded by the instructor.

Because these students were naïve to stable isotope analyses, these data analyses were kept relatively simple (but a more sophisticated tutorial for more advanced students who have successfully completed the introductory exercise is included; Supplemental File 4).

When data analyses were completed (the third day of the exercise, on personal computers that were available in the teaching laboratory), students were assigned to write a paper modelled after a peer-reviewed scientific manuscript (SLO6) including an introduction that included background literature, complete methods, a summary of the results (including graphical representations of their data), conclusions, and a literature cited section. Students were instructed to format these according to guidelines for the journal Ecology. This paper was then graded by the instructor.

A separate advanced R tutorial was tested by 1 undergraduate student who had previously completed the introductory exercise. This tutorial used package dplyr v.0.8.2 (Wickham et al.

2009) to filter and summarize data. The advanced tutorial also included the development of

Bayesian isotope mixing models using package simmr v.0.4.1 (Parnell 2019). The student agreed to act as a tester for the advanced exercise, and worked through it individually, with limited guidance from the instructor. Tasks in the advanced R notebook exercise included loading the data, extracting elements of a data frame via indexing, subsetting data frames and matrices, conversion of data frame objects to matrices or vectors, building a simmr object, running and interpreting mixing models, creating dual isotope and box and whisker plots, and estimating diet

183 proportions of target taxa. The student turned in a completed version with properly annotated answers to questions which were then evaluated by the instructor.

Results

Sampling and Stable Isotopes

A general ecology laboratory course with 21 registered students obtained 26 biological samples from the catch-and-release area of the Taylor River below Taylor Park Reservoir, CO, in

February, 2019 (Table C-1). These biological samples spanned a wide taxonomic breadth, and could be categorized under multiple trophic levels. One sample was omitted from further analysis because it did not meet the minimum size requirements, and others collected by different groups were pooled to ensure minimum size requirements were met as well as to reduce costs. Thus, a total of 19 samples were submitted for stable isotopes analyses at ISU’s SIPERG

Laboratory, and those results are provided herein (Table C-2). In brief, corrected δ13C ranged from −17.54‰ to −29.51‰, and corrected δ15N ranged from 8.41‰ to 49.26‰ (Table C-2).

Table C-2: Corrected δ13C (VPDB) and δ15N (air) values for 19 biological samples collected by students for this laboratory exercise. δ13C values are negative as a result of biological fractionation. Percent concentrations of C and N, taxonomic identifiers, and presumed trophic levels are also provided.

Sample ID Identifier Trophic Level 13C 15N %C %N G1PAANS1 Leech Secondary Consumer -23.10 8.41 40.30 9.19 G1GRAMS2, Amphipods Detritivore -19.37 8.80 40.25 8.10 G2GRAMS3, G3DEAMS5 pooled G1GRPLS3 Stoneflies Herbivore/Detritivore -26.58 9.42 50.19 10.32 G1GREPS4 Mayflies Herbivore -24.38 8.07 48.65 10.43 G1DEDES6 Wood Detritus -18.87 19.80 49.35 0.92 G4DEDES3, Wood Detritus -23.15 10.46 44.40 1.83 G5DEDES1 pooled G5PRALS2 Algae Primary Producer -24.38 9.06 22.42 3.05 G5SHTRS3 Caddis Detritivore -29.51 8.40 47.34 9.08

184

Table C-2 (continued)

Sample ID Identifier Trophic Level 13C 15N %C %N G4GREPS2, Stoneflies Herbivore/Detritivore -27.25 8.88 48.78 10.10 G5GRCHS4 pooled G2DEDES1 Wood Detritus -17.54 49.26 50.21 0.18 G2GRCHS2 Midges Detritivore -26.01 8.50 49.56 9.67 G2GREPS4 Mayflies Herbivore -26.20 7.41 45.50 9.20 G2GRPLS5 Stoneflies Herbivore/Detritivore -26.78 10.51 47.04 10.78 G2PRALS6, Algae Primary Producer -22.58 8.62 21.69 2.75 G4PRALS1 pooled G3COCHS2, Midges Detritivore -25.87 6.94 37.59 7.83 G4GRCHS4 pooled G3SHPLS1 Stoneflies Herbivore/Detritivore -27.25 9.62 49.34 10.35 G3SHEPS3 Mayflies Herbivore -26.43 7.68 49.37 9.75 G3PRALS4 Algae Primary Producer -22.75 8.50 16.44 2.22 G3APSAS6 Trout Apex Predator -27.48 11.33 29.98 8.89

Data Analyses

Students following a well-annotated R notebook tutorial (Supplemental File 3) produced several graphical representations of the data (SLO5). These included dual isotope plots (Figure

C-3), frequency histograms (Figure C-4), distance matrices (Figure C-5) and cluster plots (Figure

C-6). Students were able to manage data, generate graphical representations of their data, and answer questions pertinent to the analytical steps they were taking with high levels of success

(Table C-).

SLO Results

Student success rates were substantial for each of the student learning outcomes for a general ecology laboratory course (Table C-). While students did not achieve 100% success on all six SLOs, they did on four of the six. For the two SLOs where students achieved <100% success, students attained a 92.3% success rate on SLO3, and an 82.2% success rate on SLO6

185

(Table C-). Similarly, one student familiar with the exercise successfully completed the advanced

R exercise with a 100% success rate, and while that sample size is small and unlikely to be representative of an entire class, we are confident that students could achieve high levels of success in completing that exercise as well, either in a more advanced class, or as a follow-up exercise in a general ecology class.

Figure C-3: Student example of dual isotope plots. Left panel: Prior to removal of outliers. Right panel: Outliers removed.

186

Figure C-4: Student example of histograms using isotope data. Top panel: Full data set; Middle panel: One outlier removed; Bottom panel: Second outlier removed. Students were asked to describe the distributions, justify cut-off points for the removal of outliers, and write lines of code to remove outliers in the R tutorial. Removal of outliers was justified by the abnormal isotopic signal of two wood (detritus) samples that may have been caused by samples not being thoroughly cleaned and dried, or were contaminated by microbial activity.

187

Figure C-5: Student example of a pairwise Euclidean distance matrix comparing various trophic categories. Color coding ranges from blue (completely similar) to dark orange (highly dissimilar). Students received instruction on how to interpret these pairwise differences, then were assigned to reassess their preliminary trophic level designations.

188

Figure C-6: Student example of a cluster plot grouping samples together based on corrected δ13C and δ15N values. Students were instructed to produce numerous plots, using different numbers of clusters, then select the number of clusters they felt best represented the data and explain why in the R notebook tutorial and in their laboratory report.

189

Table C-3: Results of student learning outcomes.

SLO # Assessment Method Success Rate SLO1 Successful capture of a variety of aquatic organisms by students 100% SLO2 Determination of broad taxonomic categories by students, corrected 92.3% when necessary by the instructor (see Table 1) SLO3 Tentative assignment of samples to a trophic level by students, with 100% justification to the instructor for why (these were reassessed later in the R tutorial) SLO4 Laboratory activities supervised by the instructor and/or teaching 100% assistant SLO5 Graded R tutorial with numerous questions answered (including a 104% bonus question; Supplemental File 1) SLO6 Graded laboratory report modeled after a peer-reviewed scientific 82.2% article SLO7 Graded pre-laboratory worksheets covering stream ecology Not assessed terminology SLO8 Graded pre-laboratory worksheets covering stable isotopes Not assessed terminology

Discussion

Stable Isotopes Results

Stable isotopes are useful for assessing organismal trophic levels (Hobson, Piattt, and

Pitocchelli 1994; Vander Zanden, Casselman, and Rasmussen 1999; Hershey et al. 2017), and to match organisms to their diets (Hilderbrand et al. 1996; Richards et al. 2000; Bearhop et al.

2003; Divine et al. 2017; Patterson et al. 2019). At the conclusion of this exercise, students were able to reassess the a priori assumptions they made while categorizing the biological samples they obtained. While many of the samples were confirmed to have the trophic status they were initially thought to hold, not all were. For example, one group of students caught a leech, which they presumed to be a secondary consumer (based on the common assumption that leeches are parasites). However, the Euclidean distance matrix they produced revealed the leech’s isotopic signature to be most similar to primary producers rather than to consumers (Figure C-5).

Moreover, cluster analysis placed the leech in a group with producers, an herbivore, and a

190 detritivore (Figure C-6). Hence, the students reassessed their a priori assumption and concluded that the leech was likely a free-living nonparasitic species that fed on plant or algal material (a conclusion that is corroborated by a Bayesian isotope mixing model in the advanced R exercise;

Supplemental File 4).

Both δ13C and δ15N in living tissues vary widely across space and time. Plants, and aquatic plants in particular, are often regarded as problematic in food web studies because of a high degree of unexplained variability in their isotopic signatures (Chappuis et al. 2017). In general, plants at high elevation tend to have higher δ13C values in their tissues relative to low- elevation plants (Körner, Farquhar, and Roksandic 1988). This is due not only to the effects of lower growing season temperatures on isotopic fractionation, but also to the signal produced by low internal to external partial pressure (pCO2) ratios (Körner, Farquhar, and Roksandic 1988).

In aquatic environments, however, the signal appears to be driven largely by pH and DIC (Finlay

2001; Chappuis et al. 2017), which may be highly variable throughout the year and from location to location. Plants incorporate the DIC into their tissues and in turn serve as the ultimate source of organic carbon for the vast majority of the rest of the organisms. The absolute δ13C values of higher trophic level organisms may be shifted relative to other locations, seasons, and elevations because of the original DIC signature on which the system is based (Kobayashi et al. 2011). For this reason, ecosystems cannot be directly compared without careful site selection and controls.

Care should therefore be taken when instructing students to reference appropriate literature.

Students must be asked to compare offsets in the isotopic signatures between trophic levels in their study and others, not the absolute values, unless differences between locations are made an explicit goal of the exercise.

191

The adage ‘You are what you eat plus a few permil’ (DeNiro and Epstein 1976) has deservedly become a rule of thumb in ecosystem studies. The cumulative effect of biological isotopic fractionation is to produce trophic levels that are enriched approximately 3‰ in δ15N from producers to grazers, grazers to secondary consumers, and so forth, and enriched approximately 1‰ in δ13C from one level to the next in terrestrial systems (see Post 2002 for a thorough review of the subject). δ13C enrichment is smaller in freshwater systems, averaging

0.2‰ (France and Peters 1997). The Taylor River food web shows evidence of nitrogen enrichment with increasing trophic level, as expected, but unexpectedly shows carbon depletion between one trophic level to the next. The negative offset between the δ13C of the trout and its apparent primary food source, stoneflies, is likely to be due to time lags in tissue turnover, exacerbated by harsh winter conditions at high elevation. Fish fins are understood to have long tissue turnover times in the absence of damage, and therefore reflect diet of several months to more than a year prior to sampling, depending on the fish’s growth rate (Busst and Britton 2018).

Work by Thomas and Crowther (2015) would predict that the stoneflies which made up the trout’s diet likely have a much faster tissue turnover time than the trout, but also have more severely reduced metabolic activity in the winter months. Stoneflies sampled in the present study were likely representative of the previous fall’s food supply. This ‘time averaging’ effect is an established, though often overlooked, consideration (O’Reilly et al. 2002), and one which should be introduced to students fully. Allowing students to interact with complexity and unexpected results in a real-world context is a critical component of scientific training.

Assessment

This exercise focused on student learning over three laboratory periods via a well- organized series of field- and laboratory-based experiential learning exercises wherein students collected biological samples using a variety of methods, applied safe and effective laboratory

192 procedures to process samples, analyzed data, interpreted results, and communicated their findings.

The exercise included six student learning outcomes. SLOs 1–4 allowed students to develop new field- and laboratory-based skills. SLO5 facilitated increased confidence in students pertaining to their ability to manage and analyze data using R statistical software. SLO6 enhanced students’ scientific communication abilities. Performance assessments revealed that success rates for all six SLOs were ≥ 82.2% (Table C-). The advanced exercise was completed by an undergraduate student tester with 100% success. Therefore, students demonstrated very high levels of learning.

Educational Impact

Through these experiential learning exercises, students were able to function as ‘real’ scientists and were introduced to new field and laboratory techniques, demonstrating proficiency in the required skills to complete the exercise. While this exercise did have six SLOs, the focus was on student learning via the process of these field and laboratory activities. This exercise provided a quality educational experience that connected students to their environment in a unique way, two important components of experiential learning (Kolb 1984; Katula and

Threnhauser 1999; Kolb and Kolb 2005). Moreover, the exercise raised the students’ ecological consciousness, which may translate into improved societal, ecological, and environmental understanding (Hill, Wilson, and Watson 2004). In addition to students performing well and achieving high levels of success on the six SLOs, at the conclusion of the exercise, several students communicated verbally with the instructor that it was their favorite exercise of the semester.

Instructors interested in conducting similar activities in their own courses can view necessary supplies and costs (Supplemental File 5). Much of the necessary equipment (e.g.

193 waders, kick nets, drift nets, dissecting scopes, fume hood, drying oven, etc.) was already available for use at WCU, and glassware, consumables, and per sample costs of stable isotopes analysis on the mass spectrometer did present an ‘up front’ cost that warrants consideration. This exercise was conducted for a cost of approximately $750 in Spring Semester 2019, but subsequent activities would be less expensive (or could include more samples) with the reusable materials already purchased. As stated at the end of the introduction, this activity could easily be tailored to other ecosystems proximate to other educational institutions.

Acknowledgements

We thank Kevin Alexander for sharing knowledge of aquatic invertebrates in the Taylor

River, Holly Brunkal for sharing knowledge of Taylor River geology, Jacqueline Galang for testing the introductory and advanced R exercises and providing feedback, Justin Conover for reviewing earlier versions of R code, Miles Perry for information technology support, Kayla

Wernsing for assisting with isotope sample processing in the ISU SIPERG lab, and Nocona

Swindell for assisting with student instruction in the field and laboratory at WCU. We also thank the students of Biology 302 at Western Colorado University for their enthusiastic participation in this exercise. We are grateful to Lynn Clark and two anonymous reviewers for comments which improved the quality of the manuscript.

This work was funded entirely by student course fees; no outside funding was obtained.

Disclosure Statement

No potential conflict of interest was reported by the authors.

Data Accessibility Statement

The field sampling methods handout provided to students (Supplemental File 1), laboratory methods handout provided to students (Supplemental File 2), introductory and advanced R tutorials (Supplemental Files 3 and 4, respectively), instructor keys (Supplemental

194

Files 3Key and 4Key, respectively), materials list and budget (Supplemental File 5), a software helpfile for instructors (Supplemental File 6), and a spreadsheet of raw data used in this exercise in Microsoft Excel format (Supplemental File 7 [2019_EA_Houston_class project.xlsx]; See also

Table C-2) are available via public GitHub repository: https://github.com/hannahcarroll/Aquatic- isotopes-public

Educators who wish to teach these analyses without conducting the field work and laboratory sample preparation may use the materials as provided for their own courses. This may be particularly useful at institutions where the costs may be prohibitive.

References

Baker, J. L., M. S. Lachniet, O. Chervyatsova, Y. Asmerom, and V. J. Polyak. 2017. “Holocene Warming in Western Continental Eurasia Driven by Glacial Retreat and Greenhouse Forcing.” Nature Geoscience 10: 430. doi:10.1038/ngeo2953.

Bearhop, S., G. M. Hilton, S. C. Votier, and S. Waldron. 2004. “Stable Isotope Ratios Indicate that Body Condition in Migrating Passerines Is Influenced by Winter Habitat.” Proceedings of the Royal Society of London. Series B: Biological Sciences 271: 215– 218. doi:10.1098/rsbl.2003.0129.

Bearhop, S., R. W. Furness, G. M. Hilton, and S. C. Votier. 2003. “A Forensic Approach to Understanding Diet and Habitat Use from Stable Isotope Analysis of (Avian) Claw Material.” Functional Ecology 17: 270–275. doi:10.1046/j.1365-2435.2003.00725.x.

Boschker, H., and J. Middelburg. 2002. “Stable Isotopes and Biomarkers in Microbial Ecology.” FEMS Microbiology Ecology 40: 85–95. doi:10.1111/fem.2002.40.issue-2.

Bridgham, S. D., J. P. Megonigal, J. K. Keller, N. B. Bliss, and C. Trettin. 2006. “The Carbon Balance of North American Wetlands.” Wetlands 26: 889–916. doi:10.1672/0277- 5212(2006)26[889:TCBONA]2.0.CO;2.

Bunn, S. E., and P. I. Boon. 1993. “What Sources of Organic Carbon Drive Food Webs in Billabongs? A Study Based on Stable Isotope Analysis.” Oecologia 96: 85–94. doi:10.1007/BF00318034.

Busst, G. M. A., and J. R. Britton. 2018. “Tissue-specific Turnover Rates of the Nitrogen Stable Isotope as Functions of Time and Growth in a Cyprinid Fish.” Hydrobiologia 805: 49–60. doi:10.1007/s10750-017-3276-2.

195

Cerling, T. E. 1984. “The Stable Isotopic Composition of Modern Soil Carbonate and Its Relationship to Climate.” Earth and Planetary Science Letters 71: 229–240. doi:10.1016/0012-821X(84)90089-X.

Chappuis, E., V. Seriñá, E. Martí, E. Ballesteros, and E. Gacia. 2017. “Decrypting Stable-isotope (δ13C and δ15N) Variability in Aquatic Plants.” Freshwater Biology 62: 1807–1818.

Cummins, K. W., M. A. Wilzbach, D. M. Gates, J. B. Perry, and W. Bruce. 1989. “Shredders and Riparian Vegetation Stream Invertebrates.” BioScience 39: 24–30. doi:10.2307/1310804.

Cummins, K. W., and M. J. Klug. 1979. “Feeding Ecology of Stream Invertebrates.” Annual Review of Ecology and Systematics 10: 147–172. doi:10.1146/annurev.es.10.110179.001051.

DeNiro, M. J., and S. Epstein. 1976. “You are What You Eat (Plus a Few ‰): The Carbon Isotope Cycle in Food Chains.” Geological Society of America Abstracts with Programs 8: 834–835.

Divine, L. M., B. A. Bluhm, F. J. Mueter, and K. Iken. 2017. “Diet Analysis of Alaska Arctic Snow Crabs (Chionoecetes opilio) Using Stomach Contents and δ13C and δ15N Stable Isotopes.” Deep-Sea Research Part II: Topical Studies in Oceanography 135: 124–136. doi:10.1016/j.dsr2.2015.11.009.

Epstein, S., R. Buchsbaum, H.A. Lowenstam, and H.C. Urey. 1953. “Revised Carbonate-water Isotopic Temperature Scale.” Bulletin of the Geological Society of America 64: 1315– 1326. doi:10.1130/0016-7606(1953)64[1315:RCITS]2.0.CO;2.

Finlay, J. C. 2001. “Stable-carbon-isotope Ratios of River Biota: Implications for Energy Flow in Lotic Food Webs.” Ecology 82: 1052–1064.

France, R. L., and R. H. Peters. 1997. “Ecosystem Differences in the Trophic Enrichment of 13C in Aquatic Food Webs.” Canadian Journal of Fisheries and Aquatic Sciences 54: 1255– 1258. doi:10.1139/f97-044.

Fry, B. 2006. Stable Isotope Ecology. New York, NY: Springer New York.

Hershey, A. E., R. M. Northington, J. C. Finlay, and B. J. Peterson. 2017. “Stable Isotopes in Stream Food Webs.” In Methods in Stream Ecology: Volume 2: Ecosystem Structure, edited by G. A. Lamberti and F. R. Hauer, 3rd ed., 3–20. Cambridge, MA: Academic Press. https://doi.org/10.1016/B978-0-12-813047-6.00001-2.

Hilderbrand, G. V., S. D. Farley, C. T. Robbins, T. A. Hanley, K. Titus, and C. Servheen. 1996. “Use of Stable Isotopes to Determine Diets of Living and Extinct Bears.” Canadian Journal of Zoology 74: 2080–2088. doi:10.1139/z96-236.

Hill, S. B., S. Wilson, and K. Watson. 2004. “Learning Ecology. A New Approach to Learning and Transforming Ecological Consciousness.” In Learning toward an Ecological

196

Consciousness: Selected Transformative Practices, edited by E. V. O’Sullivan and M. M. Taylor, 47–64, New York, NY: Palgrave Macmillan US.

Hobson, K. A. 2016. “Tracing Origins and Migration of Wildlife Using Stable Isotopes : A Review.” Oecologia 120: 314–326. doi:10.1007/s004420050865.

Hobson, K. A., J. F. Piattt, and J. Pitocchelli. 1994. “Using Stable Isotopes to Determine Seabird Trophic Relationships.” Journal of Animal Ecology 63: 786–798. doi:10.2307/5256.

Hogan, J. D., M. J. Blum, J. F. Gilliam, N. Bickford, and P. B. McIntyre. 2014. “Consequences of Alternative Dispersal Strategies in a Putatively Amphidromous Fish.” Ecology 95: 2397–2408. doi:10.1890/13-0576.1.

Hsieh, J. C. C., and C. J. Yapp. 1999. “Stable Carbon Isotope Budget of CO2 in a Wet, Modern Soil as Inferred from Fe (CO3)OH in Pedogenic Goethite: Possible Role of Calcite Dissolution.” Geochimica et Cosmochimica Acta 63: 767–783. doi:10.1016/S0016- 7037(99)00062-9.

Jardine, T. D., M. A. Gray, S. M. McWilliam, and R. A. Cunjak. 2005. “Stable Isotope Variability in Tissues of Temperate Stream Fishes.” Transactions of the American Fisheries Society 134: 1103–1110. doi:10.1577/T04-124.1.

Junk, W. J., P. B. Bayley, and R. E. Sparks. 1989. “The Flood Pulse Concept in River-floodplain Systems.” In Proceedings of the International Large River Symposium, edited by D. P. Dodge, 110–127, Canadian special publication of fisheries and aquatic sciences. Ontario, Canada.

Katula, R. A., and E. Threnhauser. 1999. “Experiential Education in the Undergraduate Curriculum.” Communication Education 48(3): 238–55. https://doi.org/10.1080/03634529909379172.

Kobayashi, S., F. Akamatsu, K. Amano, S. Nakanishi, and Y. Oshima. 2011. “Longitudinal Changes in δ13C of Riffle Macroinvertebrates from Mountain to Lowland Sections of a Gravel-bed River.” Freshwater Biology 56: 1434–1446. doi:10.1111/fwb.2011.56.issue- 7.

Kolb, A. Y., and D. A. Kolb. 2005. “Learning Styles and Learning Spaces: Enhancing Experiential Learning in Higher Education.” Academy Of Management Learning & Education 4 (2): 193–212. doi: 10.5465/amle.2005.17268566.

Kolb, D. A. 1984. Experiential Learning: Experience as The Source Of Learning and Development. Englewood Cliffs, NJ: Prentice Hall.

Körner, C., G. D. Farquhar, and Z. Roksandic. 1988. “A Global Survey of Carbon Isotope Discrimination in Plants from High Altitude.” Oecologica 74: 623–632. doi:10.1007/BF00380063.

197

Lipp, J., P. Trimborn, P. Fritz, H. Moser, B. Becker, and B. Frenzel. 1991. “Stable Isotopes in Tree Ring Cellulose and Climatic Change.” Tellus 43B: 322–330. doi:10.3402/tellusb.v43i3.15280.

McDermott, F. 2004. “Palaeo-climate Reconstruction from Stable Isotope Variations in Speleothems: A Review.” Quaternary Science Reviews 23: 901–918. doi:10.1016/j.quascirev.2003.06.021.

Merritt, R., and K. Cummins. 1996. An Introduction to the Aquatic Insects of North America. 3rd ed. Dubuque, IA: Kendall/Hunt Publishing Company.

Molles, M., and A. Sher. 2019. Ecology: Concepts & Applications. 8th ed. New York, NY: McGraw Hill Education.

O’Leary, M. H. 1988. “Carbon Isotopes in Photosynthesis.” BioScience 38: 328–336. doi:10.2307/1310735.

O’Reilly, C. M., R. E. Hecky, A. S. Cohen, and P. D. Plisnier. 2002. “Interpreting Stable Isotopes in Food Webs: Recognizing the Role of Time Averaging at Different Trophic Levels.” Limnology and Oceanography 47: 306–309. doi:10.4319/lo.2002.47.1.0306.

O’Reilly, C. M., S. R. Alin, P.-D. Plisnier, A. S. Cohen, and B. A. McKee. 2003. “Climate Change Decreases Aquatic Ecosystem Productivity of Lake Tanganyika, Africa.” Nature 424: 766–768. doi:10.1038/nature01833.

Parnell, A. 2019. simmr: A stable isotope mixing model.

Patterson, D. B., D. R. Braun, K. Allen, W. A. Barr, A. K. Behrensmeyer, M. Biernat, S. B. Lehmann, et al. 2019. “Comparative Isotopic Evidence from East Turkana Supports a Dietary Shift within the Genus Homo.” Nature Ecology & Evolution 3: 1048–1056. doi:10.1038/s41559-019-0916-0.

Peterson, B. J., and B. Fry. 1987. “Stable Isotopes in Ecosystem Studies.” Annual Review of Ecology and Systematics 18: 293–320. doi:10.1146/annurev.es.18.110187.001453.

Reynolds, D. J., I. R. Hall, J. D. Scourse, C. A. Richardson, A. D. Wanamaker, and P. G. Butler. 2017. “Biological and Climate Controls on North Atlantic Marine Carbon Dynamics over the Last Millennium: Insights from an Absolutely Dated Shell-Based Record from the North Icelandic Shelf.” Global Biogeochemical Cycles 31: 1718–1735. doi:10.1002/gbc.v31.12.

Reynolds, D. J., J. D. Scourse, P. R. Halloran, A. J. Nederbragt, A. D. Wanamaker, P. G. Butler, C. A. Richardson, et al. 2016. “Annually Resolved North Atlantic Marine Climate over the Last Millennium.” Nature Communications 7: 13502. doi:10.1038/ncomms13502.

Richards, M. P., P. B. Pettitt, E. Trinkaus, F. H. Smith, M. Paunovic, and I. Karavanic. 2000. “Neanderthal Diet at Vindija and Neanderthal Predation: The Evidence from Stable

198

Isotopes.” Proceedings of the National Academy of Sciences 97:7663–7666. doi:10.1073/pnas.120178997.

Rubenstein, D. R., C. P. Chamberlain, R. T. Holmes, M. P. Ayres, J. R. Waldbauer, G. R. Graves, and N. C. Tuross. 2002. “Linking Breeding and Wintering Ranges of a Migratory Songbird Using Stable Isotopes.” Science 295: 1062–1065. doi:10.1126/science.1067124.

Rubenstein, D. R., and K. A. Hobson. 2004. “From Birds to Butterflies: Animal Movement Patterns and Stable Isotopes.” Trends in Ecology and Evolution 19: 256–263. doi:10.1016/j.tree.2004.03.017.

Sharp, Z. 2007. Principles of Stable Isotope Geochemistry. 1st ed. Saddle River, NJ: Pearson/Prentice Hall.

Steinbeiss, S., G. Gleixner, and M. Antonietti. 2009. “Effect of Biochar Amendment on Soil Carbon Balance and Soil Microbial Activity.” Soil Biology and Biochemistry 41: 1301– 1310. doi:10.1016/j.soilbio.2009.03.016.

Team, R. D. C. 2015. “R: A Language and Environment for Statistical Computing.” R Foundation for Statistical Computing. Vienna, Austria.

Thomas, S. M., and T. W. Crowther. 2015. “Predicting Rates of Isotopic Turnover across the Animal Kingdom: A Synthesis of Existing Data.” Journal of Animal Ecology 84: 861– 870. doi:10.1111/1365-2656.12326.

Thorp, J. H., and M. D. Delong. 1994. “The Riverine Productivity Model: An Heuristic View of Carbon Sources and Organic Processing in Large River Ecosystems.” Oikos 70: 305– 308. doi:10.2307/3545642.

Vander Zanden, M. J., J. M. Casselman, and J. B. Rasmussen. 1999. “Stable Isotope Evidence for the Food Web Consequences of Species Invasions in Lakes.” Nature 401: 464–467. doi:10.1038/46762.

Vannote, R. L., G. W. Minshall, K. W. Cummins, J. R. Sedell, and C. E. Cushing. 1980. “The River Continuum Concept.” Canadian Journal of Fisheries and Aquatic Sciences 37: 130–137. doi:10.1139/f80-017.

Wanamaker, A. D., K. J. Kreutz, H. W. Borns, D. S. Introne, S. Feindel, S. Funder, P. D. Rawson, and B. J. Barber. 2007. “Experimental Determination of Salinity, Temperature. Growth, and Metabolic Effects on Shell Isotope Chemistry of Mytilus edulis Collected from Maine and Greenland.” Paleoceanography 22: 1–12. doi:10.1029/2006PA001352.

Wassenaar, L., and K. Hobson. 1998. “Natal Origins of Migratory Monarch Butterflies at Wintering Colonies in Mexico: New Isotopic Evidence.” Proceedings of the National Academy of Science 95: 4. doi:10.1073/pnas.95.26.15436.

199

West, J. B., G. J. Bowen, T. E. Cerling, and J. R. Ehleringer. 2006. “Stable Isotopes as One of Nature’s Ecological Recorders.” Trends in Ecology and Evolution 21: 408–414. doi:10.1016/j.tree.2006.04.002.

Whitney, N. M., B. J. Johnson, P. T. Dostie, K. Luzier, and A. D. Wanamaker. 2019. “Paired Bulk Organic and Individual Amino Acid δ15N Analyses of Bivalve Shell Periostracum: A Paleoceanographic Proxy for Water Source Variability and Nitrogen Cycling Processes.” Geochimica Et Cosmochimica Acta 254: 67–85. doi:10.1016/j.gca.2019.03.019.

Wickham, H., R. Francois, L. Henry, and K. Muller. 2009. Ggplot2: Elegant Graphics for Data Analysis. New York, NY: Springer-Verlag.

Wood, S. 2006. Generalized Additive Models: An Introduction with R. C and Hall. London.