1 Geochemistry of Soils from the Shackleton Glacier Region

Home , Antarctic Plateau, Ashworth Glacier, Law Glacier, Perez Glacier, Shackleton Glacier, Souchez Glacier

Geochemistry of soils from the Shackleton Glacier region, Antarctica, and implications

for glacial history, salt dynamics, and biogeography

Dissertation

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy

in the Graduate School of The Ohio State University

Melisa A. Diaz

Graduate Program in Earth Sciences

The Ohio State University

2020

Dissertation Committee

W. Berry Lyons, Advisor

Thomas Darrah

Elizabeth Griffith

Bryan Mark

Copyrighted by

Melisa A. Diaz

2020

Abstract

Though the majority of the Antarctic continent is covered by ice, portions of

Antarctica, mainly in the McMurdo Dry Valleys and the Transantarctic Mountains

(TAM), are currently ice-free. The soils which have developed on these surfaces have been re-worked by the advance and retreat of glaciers since at least the Miocene. They are generally characterized by high salt concentrations, low amounts of organic carbon, and low soil moisture in a polar desert regime. Ecosystems that have developed in these soils have few taxa and simplistic dynamics, and can therefore help us understand how ecosystems structure and function following large-scale changes in climate, such as glacial advance and retreat. During the 2017-2018 austral summer, 220 surface soil samples and 25 depth profile samples were collected from eleven ice-free areas along the

Shackleton Glacier, a major outlet glacier of the East Antarctic Ice Sheet. A subset of 27 samples were leached at a 1:5 soil to deionized (DI) water ratio and analyzed for stable

- 15 17 2- 34 18 isotopes of water-soluble NO3 (δ N and Δ O) and SO4 (δ S and δ O), and seven

13 18 soils were analyzed for δ C and δ O of HCO3 + CO3 to understand the sources and cycling of salts in TAM soils. The depth profiles and a subset of surface soil (21) samples

10 - were analyzed for concentrations of meteoric Be and/or water-soluble NO3 to estimate relative surface exposure ages along the length of the Shackleton Glacier. Finally, water- soluble ion data from all 220 samples were correlated with geography and

ii geomorphology to elucidate geochemical trends and gradients. The relationship between geochemistry and geography was further used to predict/estimate geochemical gradients in the TAM using interpolation and machine learning techniques. Collectively, these measurements and data show that atmospheric deposition is an important source of water- soluble anions, which have possibly been accumulating in some soils since at least the

Pleistocene. Soils near the terminus of the Shackleton Glacier and near the glacier margins are younger than those further south towards the Polar Plateau and further inland. These soils have nutrient and salt concentrations most suitable for soil invertebrates. The geochemistry of TAM soils is related to geography, glacial history, and the present and past availability of liquid water, and these parameters can be used to effectively predict/model geochemical gradients. These findings will greatly aid in our collective understanding of habitat suitability and past refugia in Antarctic terrestrial systems, and may help predict how ecosystems will respond to future climate transitions.

iii

Dedication

To my mom and grandma

Acknowledgments

The land on which I have based my MS and PhD education for the last 4+ years has long served as a site of meeting and exchange amongst Indigenous peoples, specifically the Shawnee, Miami, Wyandot and Delaware Nations. I honor and respect the diverse Indigenous peoples connected to this territory.

I am grateful to the United States Antarctic Program (USAP), Antarctic Science

Contractors (ASC) and Petroleum Helicopters Inc. (PHI) for logistical and field support; otherwise, the samples analyzed in this work would not exist. This work was supported by a National Science Foundation (NSF) OPP grant (1341631) awarded to WBL, NSF

GRFP fellowship (60041697) awarded to MAD, International Association of

Geochemistry (IAGC) Research Grant awarded to MAD, Geological Society of America

(GSA) Student Research Grant awarded to MAD, and a Purdue SEED proposal awarded to MAD. Additional financial support from The Ohio State University awarded to MAD includes Friends of Orton Hall (FOH), the Women’s Place, and the Ray Travel Award. I am especially grateful to Lisa Rom (NSF) for the years of financial support for outreach from OPP. Many collaborators have assisted in the analysis of these samples. I thank Dan

Gilbert, BS 2018, for his help in leaching the soils and measuring the water content. Drs.

Greg Michalski at Purdue University, Anna Szynkiewicz at the University of Tennessee

Knoxville, Robert Gregory at Southern Methodist University, Paul Bierman and Lee

Corbett at the University of Vermont, and W. Andrew Jackson at Texas Tech were instrumental in all isotopic and perchlorate/chlorate measurements.

An African proverb states, “It takes a village to raise a child.” My village is grand and extends through several colleges, centers, and organizations at The Ohio State

University, to domestic and international universities, to funding agencies, and finally to my friends, family, and loved ones.

My friends, family, and loved ones have been an incredible emotional support throughout this journey. Without them, my hair would surely have turned 100% grey years ago. Though I often called them out of the blue with heated or excited rambles, my mom and grandma simply listened and expressed how proud they were of me. Their love meant the world to me, especially during difficult times. I am the woman I am today because of them and lean on their strength. This document, the culmination of years of hard work, is for them.

Even though we’ve been apart for a while, I thank Alicia DeGiso for being the greatest friend for the last 15 years. I never would have thought when we met in 7th grade that we would be inseparable and that we would support each other through doctoral programs. See you soon, Dr. DeGiso. I thank Melissa Wrzesien, Kelsey Danner, Allison

Chartrand, Demie Huffman, Mary Knights, and Natalie Browning for being such wonderful people and life-long friends. I will miss our girl’s nights and “book club” but look forward to our future adventures. Thank you to all my close SES friends, Casey

Saup, Emma Oti, Myles Moore, and Joe Schulze, for all the smiles, laughter, and of course, complaining. I will especially miss “Brown Lunch” with dear friends Nlingi

Habana and Deon Knights. I thank Cole Bradley for always smiling through my grumpy moods and for teaching me how to have a little more fun. He is my biggest fan and I am grateful for his love and support.

My first academic family was the Lyons Group, and the members are so much more to me than just scientific trainers. The brilliant sister duo, Sue and Kathy Welch, are role models and friends. Their knowledge of geochemistry is unparalleled, and I have learned so much about analytical chemistry from them since I joined the group. Kathy and I had numerous great conversations about not just science, but anything and everything from personality types to fantastic bread. I thank her for teaching me all about

McMurdo-isms. Sue and I bonded over our love of stouts and looking at beautiful euhedral salt crystals under the scanning electron microscope. I will miss wine nights and candy binges. Chris Gardner and Adolfo Calero are like older and younger brothers. Of course we occasionally bickered and annoyed each other, but our time together at New

Harbor Camp was the perfect final field season to my student career. Adolfo is the one of the sweetest, kindest, and funniest people I have met. He was always eager to help me in any way he could, even if it was just listening to me reading sentences aloud over and over to see which made the most sense. Chris taught me patience and flexibility both in my professional and academic life. Even if he was busy and I rushed into his office without warning with a flurry of questions, he would listen and offer his help and guidance. I will miss our walks, lunches, and silly conversations. And of course, I am

vii grateful he broke his leg, allowing me to take his spot on the Shackleton Glacier expedition. Thanks, Chris!

I express my gratitude to Drs. Byron Adams and Marci Shaver-Adams (Brigham

Young University), Diana Wall (Colorado State University), Ian Hogg (Polar Knowledge

Canada; University of Waikato), and Noah Fierer (University of Colorado Boulder).

They have served as mentors, coaches, and fans throughout the last three years and I am so grateful to have had their support. Their guidance in both the field and in research has greatly shaped my scientific growth and perspective. The two weeks we spent together visiting some of the most beautiful features on Earth are among my fondest memories.

Despite being the youngest member on the team, they regarded me as a colleague, valued my opinions and expertise, and invested in my development. Three years ago, I knew little of Antarctic biology. Now, ecology is a regular word in my vocabulary, thanks to them.

The Byrd Polar and Climate Research Center was a shining beacon in my studies.

Byrd felt like home. Michele Cook and Charmaine Koch are fantastic and powerful individuals. They taught me much about the financial side of research and we bonded over our love of animals. I thank Jason Cervenec for helping me develop my scientific communication and outreach skills. I still have a long way to go, but I have learned so much from him. I am thankful for the community and comradery of center members that

I believe all departments strive for.

The Society for the Advancement of Chicanos/Hispanics and Native Americans in

Science (SACNAS) and the Office of Diversity and Inclusion (ODI) supplemented my

viii education in the most meaningful ways. Yolanda Zepeda and Marcela Hernandez were two of the best mentors I could have asked for. With each step forward, they were always there to guide me if I was ever unsure or uncertain. They helped provide a community of support and did everything in their power (sometimes beyond!) to make sure all of their students were cared for. Through SACNAS and ODI, I was able to maintain my ties to my heritage and create ties with the greater Columbus community. My outreach and education experiences with these groups shaped my career and perspective in higher education. They helped me find my voice in addressing issues concerning diversity, inclusion, and equity. SACNAS was an oasis in graduate school and I am thankful to

Steven Villanueva, Miguel Lopez, Priscila Rodriguez-Garcia, Ally Langley, and Juan

Barajas for being fun and inspiring scientists and friends. The National SACNAS

Conference, OSU SACNAS chapter meetings, weekly visits to high schools through

Latinx Space for Enrichment and Research (LASER), and the fact that I knew I always had a home away from home at ODI helped me maintain a sense of optimism and determination that propelled me through my degree.

I thank my committee members, Drs. Tom Darrah, Liz Griffith, Joel Barker, and

Bryan Mark for their guidance, expertise, and patience. They always had meaningful and thoughtful insight during committee meetings, which helped my projects deepen and grow.

Finally, words cannot express how grateful I am to Berry Lyons, my MS and PhD advisor. About 5 years ago, I sent Berry an email out of the blue asking if he was looking for graduate students. He could have easily dismissed me, as the application deadline had

ix passed over 5 months prior and I had struggled with my GPA the first two years of my undergraduate education. Instead, Berry listened and heard my passion for research. He hired me as a research associate, then took me on as a MS student, and finally encouraged me to pursue a PhD. I know that I can be a handful sometimes, but Berry was caring and understanding. He has a unique way of both offering critique and encouragement at the same time. He is the most creative scientist I know, and I am fascinated by his ability to carve impactful stories seemingly out of nowhere. He encouraged me to pursue my own research creativity, even if it meant trying something outside our wheelhouse. He struck a perfect balance between letting me run wild and guiding me. Berry understood how important diversity and inclusion is to me and supported me in any way he could. He even became a SACNAS member! Over the years, we learned much about each other and developed a mutual respect. I am the scientist I am today because of him and will forever be grateful for all he has done.

I can continue with more acknowledgments and gratitude, but with a glance at the page count, I already feel bad for the trees. Many thanks to everyone once again!

P.S. Thank you to Sirius Lee Black and Solana Renée for being the cutest and loudest cats.

Vita

2017-2020 National Science Foundation Graduate Research Fellow, The Ohio State University, School of Earth Sciences

2017 M.S. in Earth Sciences, The Ohio State University

2016-2017 Graduate Research Associate and Graduate Teaching Associate, The Ohio State University

2015 Research Associate, Lyons Group, The Ohio State University

2014 B.S. in Earth and Environmental Sciences, University of Rochester Minors in Chemistry and Japanese, University of Rochester

Publications

Diaz, M.A., Jackson, W.A., Adams, B.J., Wall, D.H., Hogg, I.D., Fierer, N., Welch, S.A., Gardner, C.B., Lyons, W.B. (In preparation). Geochemical zones and environmental gradients for soils from the Central Transantarctic Mountains, Antarctica. Antarctic Science.

Lyons, W.B., Carey, A.E., Gardner, C.B., Welch, S.A., Smith, D.F., Szynkiewicz, A., Diaz, M.A., Croot, P., Henry, T., Flynn, R. (Submitted). The Geochemistry of Irish Rivers. Geochimica et Cosmochimica Acta.

Diaz, M.A., Fortner, S.K., Lyons, W.B. (In revision). High resolution concentration- discharge relationships in managed watersheds: a 30+ year analysis.

Dragone, N.B., Diaz, M.A., Hogg, I., Lyons, W.B., Jackson, W.A., Wall, D.H., Adams, B.J., Fierer, N. (In revision). Exploring the boundaries of microbial habitability in soil. Proceedings of the National Academy of Sciences.

Diaz, M.A., Bierman, P.R., Corbett, L.B., Adams, B.J., Hogg, I., Fierer, N., Wall, D.H., Lyons, W.B. (2020; in review). Relative terrestrial exposure ages inferred from meteoric- 10 - Be and NO3 concentrations in soils along the Shackleton Glacier, Antarctica. Earth Surf. Dynam. Discuss., https://doi.org/10.5194/esurf-2020-50.

Diaz, M.A., Li, J., Michalski, G., Adams, B.J., Wall, D.H., Hogg, I., Fierer, N., Gardner, C.B., Lyons, W.B. (Accepted). Stable isotopes of nitrate, sulfate, and carbonate in soils from the Transantarctic Mountains, Antarctica: A record of atmospheric deposition and chemical weathering. Frontiers in Earth Science special issue: Novel Isotope Systems and Biogeochemical Cycling During Cryospheric Weathering in Polar Environments.

Diaz, M.A., Welch, S.A., Sheets, J., Welch, K.A., Adams, B.J., Khan, A.L., McKnight, D.M., Cary, S.C., Lyons, W.B. (2020). Geochemistry of Aeolian Material from the McMurdo Dry Valleys, Antarctica: Insights into Dust Sources. Earth and Planetary Science Letters 547, 116460. https://doi.org/10.1016/j.epsl.2020.116460.

Lyons, W.B., Foley, K.K., Carey, A.E. Diaz, M.A., Bowen, G.J., Cerling, T. (2020). The Isotopic Geochemistry of CaCO3 Encrustations in Taylor Valley, Antarctica: Implications for their Origin. Acta Geographica Slovenica 60(1): 105-119.

Diaz, M.A., Lyons, W.B., Adams, B.J., Welch, S.A., Khan, A.L., McKnight, D.M., Cary, S.C. (2018). Major Oxide Chemistry of Mineral Dust, McMurdo Dry Valleys, Antarctica: Revisited. Proscience 5: 25-30. DOI:10.14644/dust.2018.005

Diaz, M.A., Adams, B.J., Welch, K.A., Welch, S.A., Opiyo, S., Khan, A.L., McKnight, D.M., Cary, S.C., Lyons, W.B. (2018). Aeolian Material Transport and its Role in Landscape Connectivity in the McMurdo Dry Valleys, Antarctica. Journal of Geophysical Research: Earth Surface. DOI: 10.1029/2017JF004589.

Diaz, M.A., Yu, H., Deuerling, K.M., Wörner, G., Gardner, C.B., Harmon, R.S., Goldsmith, S.T., Carey, A.E., Lyons, W.B. (2017). The Flux of Saharan Dust to Panama and its Influence on Soil Geochemistry. Proscience 3: 38-43. DOI:10.14644/dust.2016.006

Fields of Study

Major Field: Earth Sciences

xii

Table of Contents

Abstract ...... ii Dedication ...... iv Acknowledgments ...... v Vita ...... xi Table of Contents ...... xiii List of Tables...... xvii List of Figures ...... xviii Chapter 1. Introduction ...... 1 1.1. Ice-free areas in Antarctica ...... 1 1.2. Brief overview of Antarctic glacial history ...... 3 1.3. Antarctic terrestrial ecosystems ...... 4 1.4. The Shackleton Glacier region ...... 5 1.5. Chapter overviews ...... 6 Chapter 2. Stable isotopes of nitrate, sulfate, and carbonate in soils from the Transantarctic Mountains, Antarctica: A record of atmospheric deposition and chemical weathering ...... 16 2.1. Abstract ...... 16 2.2. Introduction ...... 18 2.3. Methods ...... 20 2.3.1. Study site ...... 20 2.3.2. Sample collection ...... 21 2.3.3. Water-soluble leaches...... 22 2.3.4. Major ions ...... 22 2.3.5. Nitrogen and oxygen isotope analysis of nitrate ...... 23 2.3.6. Sulfur and Oxygen isotope analysis of sulfate...... 24 2.3.7. Carbon and oxygen isotope analysis of carbonate ...... 25

xiii

2.3.8. Scanning Electron Microscopy ...... 25 2.4. Results ...... 26 2.4.1. Major ion concentrations ...... 26 2.4.2. Trends in salt distributions ...... 26 15 17 - 2.4.3. δ N and Δ O of NO3 ...... 27 2.4.4. δ34S and δ18O in sulfate ...... 28 2.4.5. Inorganic δ13C and δ18O of carbonate ...... 28 2.5. Discussion ...... 29 2.5.1. Water-soluble salt compositions ...... 29 - 2.5.2. Stratospheric and photochemical processes as sources of NO3 ...... 31 - 2.5.3. Primary and secondary atmospheric, and chemical weathering derived SO4 . 37 2.5.4. Cryogenic carbonate mineral formation and isotope equilibrium ...... 42 2.6. Conclusions ...... 45 10 - Chapter 3. Relative terrestrial exposure ages inferred from meteoric Be and NO3 concentrations in soils along the Shackleton Glacier, Antarctica ...... 69 3.1. Abstract ...... 69 3.2. Introduction ...... 70 3.3. Background...... 71 3.3.1. Stability of the EAIS ...... 71 3.3.2. Cosmogenic nuclide exposure age dating and meteoric 10Be systematics ...... 73 3.4. Study sites...... 74 3.5. Methods ...... 76 3.5.1. Sample collection ...... 76 3.5.2. Analytical methods ...... 77 3.5.3. Exposure age model ...... 78 3.6. Results ...... 81 3.6.1. Surface concentrations of meteoric 10Be and grain size ...... 81 3.6.2. Calculated maximum and inheritance-corrected exposure ages ...... 82 3.6.3. Estimated exposure ages for sites without meteoric 10Be depth profiles ...... 82 3.7. Discussion ...... 84 3.7.1. Calculated and estimated exposure age validation ...... 84 - 3.7.2. NO3 as an efficient exposure age dating tool ...... 86 3.7.3. Implications for ice sheet dynamics ...... 88 xiv

3.8. Conclusions ...... 90 Chapter 4. Geochemical zones and environmental gradients for soils from the Central Transantarctic Mountains, Antarctica ...... 105 4.1. Abstract ...... 105 4.2. Introduction ...... 106 4.3. Study Sites ...... 109 4.4. Methods ...... 110 4.4.1. Sample collection and preparation ...... 110 4.4.2. Analytical analysis of water-soluble anions, cations, and nutrients...... 111 4.4.3. Data interpolation and machine learning ...... 112 4.5. Results ...... 114 4.5.1. Geochemistry of upper, middle, and lower glacier zones ...... 114 4.5.2. Statistical geochemical variability ...... 116 4.5.3. Interpolation and machine learning model performance ...... 117 4.6. Discussion ...... 119 4.6.1. Implications for ecological habitat suitability ...... 119 4.6.2. Machine learning as a tool to predict soil geochemical trends ...... 123 4.7. Conclusions ...... 126 Chapter 5. Conclusions ...... 140 References ...... 145 Appendix A. Supplementary materials for Chapter 3 ...... 162 Figure A1. Max meteoric 10Be concentration versus 10Be inventory ...... 163 Table A1. Soil grain size distribution of surface samples and depth profiles from Roberts Massif, Bennett Platform, and Thanksgiving Valley...... 164 - 10 Table A2. NO3 concentrations and estimate of Be concentration from linear - 10 relationship between NO3 and Be...... 165 Appendix B. Supplementary materials for Chapter 4...... 166 Table B1. Analytical precision and accuracy for water-soluble salts...... 167 Table B2. Water-soluble salt concentrations from 1:5 soil to water leaches from the Shackleton Glacier region...... 168 Table B3. Random forest model imputed training dataset...... 200 Table B4. Random forest model imputed testing dataset...... 227 Figure B1. Images of biology at Shackleton Glacier terminus ...... 233

Code B1. R code for multiple linear regression and random forest models ...... 234

xvi

List of Tables

Table 2.1. Salt isotope sample geographic information and concentrations of water- soluble ions in soil leaches ...... 48 15 17 - 34 18 2- Table 2.2. δ N and Δ O of NO3 and δ S and δ O of SO4 ...... 53 Table 2.3. δ13C and δ18O of total inorganic carbon (TIC) in bulk soil samples ...... 55 Table 2.4. δ15 Two-component mixing model ...... 56 Table 2.5. Sulfur sequential extractions on bulk sediment ...... 58 Table 2.6. δ34S and δ18O three-component mixing model ...... 59 Table 3.1. Meteoric 10Be sampling locations ...... 92 Table 3.2. Meteoric 10Be concentrations ...... 93 - Table 3.3. Calculated and NO3 estimated exposure ages...... 95 Table 3.4. Estimated exposure ages from maximum 10Be concentration ...... 96 Table 4.1. Overview of geography, soil moisture, and water-soluble ions ...... 128 Table 4.2. Multiple linear regression and random forest model statistics ...... 129 Table 4.3. Multiple linear regression and random forest statistics between predicted and measured concentrations ...... 130 Table 4.4. Multiple linear regression and random forest model (MAE) and (RMSE) .... 132

xvii

List of Figures

Figure 1.1. Map of the Transantarctic Mountains ...... 8 Figure 1.2. Shackleton Glacier surface ...... 9 Figure 1.3. Most and least visited terrestrial environments in Antarctica ...... 10 Figure 1.4. Recent glacial and interglacial periods from EPICA Dome C ice core ...... 11 Figure 1.5. Modeled ice sheet height during the Last Glacial Maximum ...... 12 Figure 1.6. Antarctic nematode habitat suitability model ...... 13 Figure 1.7. Map of the Shackleton Glacier located in the central Transantarctic Mountains ...... 14 Figure 1.8. Aerial photo of the Shackleton Glacier Camp (85° S, 176° W) ...... 15 Figure 2.1. Study sites for salt isotope analysis ...... 61 Figure 2.2. Molar ratios of salts with distance from coast ...... 62 Figure 2.3. Stable isotopes of δ15N and Δ17O in nitrate and δ34S and δ18O in sulfate ...... 63 - 2- Figure 2.4. Salt dissolution diagrams for the major NO3 and SO4 salts ...... 64 Figure 2.5. Scanning electron microscopy (SEM) backscatter emission (BSE) images of salt encrustations from Schroeder Hill ...... 65 - Figure 2.6. Isotopic composition of NO3 for Shackleton Glacier soils compared to potential sources and other hyper-arid deserts ...... 66 2- Figure 2.7. Isotopic composition of SO4 for Shackleton Glacier soils compared to potential sources and other hyper-arid deserts ...... 67 2- Figure 2.8. Isotopic composition of CO3 for Shackleton Glacier soils compared to potential sources and other hyper-arid deserts ...... 68 Figure 3.1. Overview of Shackleton Glacier region meteoric 10Be sampling locations .... 97 Figure 3.2. Sirius Group and glacial moraines ...... 98 Figure 3.3. Conceptual diagram of meteoric 10Be accumulation in Shackleton Glacier soils ...... 99 Figure 3.4. Spatial distribution of surface meteoric 10Be ...... 100 Figure 3.5. Concentration of meteoric 10Be with elevation and distance from coast ...... 101 Figure 3.6. Grain size of soil pits from Roberts Massif, Bennett Platform, and Thanksgiving Valley...... 102 10 - Figure 3.7. Depth profiles of meteoric Be and NO3 ...... 103 Figure 3.8. Estimated maximum exposure age versus distance from the coast and elevation ...... 104 Figure 4.1. Map of 220 soil sample locations in the Shackleton Glacier region ...... 133 - - Figure 4.2. Total water-soluble salts, water-soluble N:P molar ratio, and ClO4 and ClO3 compared to elevation, distance from the coast, and distance from the glacier ...... 134 Figure 4.3. Anion and cation ternary diagrams for water-soluble salts ...... 135 xviii

Figure 4.4. Principal component analysis (PCA) biplot with all anions, cations, nutrients, and soil moisture ...... 136 Figure 4.5. Spearman’s rank correlation matrix ...... 137 Figure 4.6. Inverse distance weighted (IDW) interpolations of total salt concentration . 138 Figure 4.7. R2 values for the multiple linear regression and random forest model predicted and measured values ...... 139

xix

Chapter 1. Introduction

1.1. Ice-free areas in Antarctica

Antarctica contains 70% of the world’s freshwater and ~98% of its surface is covered by ice, mainly the East and West Antarctic Ice Sheets (EAIS and WAIS, respectively). The remaining 2% of the continent is ice-free. The McMurdo Dry Valleys

(MDV), located in southern Victoria Land, constitute the largest collective ice-free area with 4,500 km2 of exposed surface (Levy, 2013). The MDV are primarily ice-free as a result of low annual precipitation and the impediment of ice flow from the EAIS by the

Transantarctic Mountains (TAM), and have low mean annual temperatures (Doran et al.,

2002; Fountain et al., 2010). The US National Science Foundation has supported the

McMurdo Long-Term Ecological Research network (MCM-LTER), located in the MDV, since 1992. Extensive research in the MDV has resulted in a breadth of knowledge regarding soil development and geochemistry, glacial history, and ecosystem structure and function in extreme terrestrial environments (Prentice et al., 1993; Freckman and

Virginia, 1998; Bockheim, 2002; Adams et al., 2006; Barrett et al., 2006; Bockheim and

McLeod, 2013; Convey et al., 2014). Ice-free areas also exist in the TAM, and can aid in our understanding of how organisms and ecosystems have persisted in Antarctica for perhaps millions of years.

The TAM traverse the Antarctic continent and extend over 3,500 km from northern Victoria Land along the Ross Sea to Coats Land near the Weddell Sea, separating the EAIS from the WAIS (Fig. 1.1). Several mountain peaks rise above the modern ice surface of the EAIS (Fig. 1.2), the highest being Mt. Kirkpatrick at 4,528 m.a.s.l., and soils develop abiotically on exposed surfaces (Bockheim, 1990, 2002). Some soils host simple ecosystems and serve as records of the complex climate of Antarctica through time (e.g. Marchant and Denton, 1996; Virginia and Wall, 1999). One of the most intriguing questions for TAM systems concerns the abundance of genetic data on soil invertebrates. Such data indicate that organisms have survived in TAM soils for millions of years, despite several glaciations throughout the Pleistocene (Stevens and

Hogg, 2003; Convey et al., 2008; Fraser et al., 2012). It is still unclear how and where these organisms found suitable glacial refugia given the high salt concentrations in high- elevation TAM soils (Lyons et al., 2016). Questions still remain regarding the sources of salts, glacial history on a local scale, and how geochemical gradients may inform habitat suitability.

Compared to the MDV and ice-free areas along the edges of the Antarctic continent, the TAM are considerably less visited and sampled (Pertierra et al., 2017) (Fig.

1.3). As a result, they are also less understood. Antarctic researchers have developed a

“roadmap” for the future of Antarctic and Southern Ocean science for the next two decades (Kennicutt et al., 2015). As one of the few remaining environments on Earth with little to no human modification, TAM soils are crucial in addressing the most pressing scientific questions. In particular, these systems can aid in our understanding of 2

how ecosystems have responded to past climatic events. This information could be used to predict how ecosystems might respond in future climate scenarios, as well as how ecosystems might have existed or persisted in extra-terrestrial soils.

1.2. Brief overview of Antarctic glacial history

Antarctica is believed to have maintained a persistent ice sheet since potentially the Eocene epoch, and these ice sheets have waxed and waned since at least the Miocene

(Gasson et al., 2016; Gulick et al., 2017). Sediment core records collected from the Ross

Sea and ice cores from the Antarctic interior show that the EAIS and WAIS have undergone several glacial and interglacial cycles (Fig. 1.4) (Augustin et al., 2004;

Talarico et al., 2012). These cycles are driven by variations in Earth’s climate (e.g. atmospheric CO2 levels), as well as variations in Earth’s orbit, known as Milankovitch cycles.

The EAIS and WAIS were at their most recent greatest extent during a period known as the Last Glacial Maximum (LGM) (~22,000 yrs. ago) (Clark et al., 2009). The

WAIS is a marine-terminating ice sheet with a grounding line below sea level, which decreases the stability of the WAIS and results in more rapid ice sheet advance and retreat when compared to the EAIS (Pollard and DeConto, 2009). As a result, the WAIS greatly expanded during the LGM across the Ross Embayment towards the coast of

Victoria Land (McKay et al., 2008; Pollard and DeConto, 2009). The EAIS is grounded above sea level and is generally more stable than the WAIS. During the LGM, the EAIS expanded along its margins and the greatest increases in height were along outlet glaciers,

which flow through exposed peaks of the TAM and drain into the Ross and Weddell Seas

(Fig. 1.5) (Anderson et al., 2002; Golledge et al., 2012; Mackintosh et al., 2014).

1.3. Antarctic terrestrial ecosystems

The exposed terrestrial environments in Antarctica include some of the least hospitable soils on Earth due to the arid polar climate, high salt concentrations, lack of sunlight in the winter, and nutrient limitations (Freckman and Virginia, 1998; Courtright et al., 2001; Barrett et al., 2007). The ecosystems that do exist are simple, with few trophic levels, few taxa, and with organisms ranging from bacteria to soil invertebrates

(Freckman and Virginia, 1998; Adams et al., 2006). These ecosystems are likely supported by modern fluxes of chemicals and nutrients from chemical weathering, atmospheric deposition, and biotic interactions, as well as by legacy carbon and nutrients from warmer and wetter periods occurring in the past (Fig. 1.6) (Moorhead et al., 1999;

Barrett et al., 2005).

Much work has focused on understanding the structure and function of ecosystems in the MDV and TAM. The simple structure allows ecologists to study fundamental questions regarding species interactions and their responses to geochemical, nutrient, and hydrologic changes in the environment (e.g. Adams et al., 2006; Courtright et al., 2001; Freckman and Virginia, 1998; Richter et al., 2014). By understanding how communities structure, function, and adapt in Antarctic soils, we can develop a better understanding of the biotic, geochemical, and environmental drivers behind ecological habitat suitability, especially in extreme terrestrial systems (Courtright et al., 2001;

Barrett et al., 2006; Magalhães et al., 2012; Bottos et al., 2020). 4

1.4. The Shackleton Glacier region

The Shackleton Glacier (85° S, 176° W), named after the prolific Antarctic explorer Sir Ernest Shackleton, is a major outlet glacier of the EAIS located to the east of the Beardmore Glacier in the central Transantarctic Mountains (CTAM) (Fig. 1.7). The glacier flows south to north and drains into the Ross Sea to help form the Ross Ice Shelf

(RIS). Peaks of the Queen Maud Mountains rise above the ice on either side of the glacier and range in elevation from ~150 m near the RIS to >3,500 m further inland. The geologic basement consists of gneiss, schist, slate, and quartzite which were intruded by granitoid batholiths in the Ross Orogeny. Devonian to Triassic rocks of the Beacon

Supergroup that have been cut by dolerite/basaltic sills overlie the basement (Elliot and

Fanning, 2008). Near the RIS, the exposed surfaces are primarily comprised of metamorphic and igneous rocks, while the Beacon Supergroup and Ferrar Dolerite are more abundant towards the Polar Plateau. These rocks serve as weathering parent materials for soil production, where the dominant processes are salinization, rubification

(soil reddening due to iron oxidation), and desert pavement formation (Bockheim, 1990).

An interdisciplinary camp (85.1°S, 175.4°W) was established on the Shackleton

Glacier for three austral summer seasons from 2015-2018 (Fig. 1.8). During the final

2017-2018 season, seven science groups were supported. The Shackleton Glacier was selected as the study site for this work for several reasons:

1) The Shackleton Glacier region has several ice-free areas with different glacial

histories.

2) The eleven exposed peaks of the Queen Maud Mountains which were sampled

allow for reproducibility across elevational and latitudinal transects, serving as

natural “replicates” for analysis.

3) Similar to the MDV, the ecosystem dynamics are simple, so ecosystem

structure and function are easier to measure and understand.

4) The soils are among the least human modified terrestrial systems on Earth.

1.5. Chapter overviews

As stated in Section 1.3, geochemistry, geology, and geography play an important role in both biodiversity and ecosystem habitat suitability for Antarctic soils. As ecological questions have become more complex, so has the need for diverse, interdisciplinary research teams. The studies summarized below are a subset of a larger

NSF study titled, “The Role of Glacial History on the Structure and Functioning of

Ecological Communities in the Shackleton Glacier Region of the Transantarctic

Mountains.”

Chapter 2, “ Stable isotopes of nitrate, sulfate, and carbonate in soils from the

Transantarctic Mountains, Antarctica: A record of atmospheric deposition and chemical weathering,” addresses a unique component of Antarctic soils: the salt content. Antarctic soils are among the most saline terrestrial systems on Earth. The concentrations of total water-soluble salts are at levels which are potentially toxic for soil invertebrates. The production and cycling of salts in TAM soils is understudied, particularly with respect to the interplay between atmospheric deposition and chemical weathering. Stable isotopes

- 15 17 2- 34 18 13 18 of NO3 (δ N and Δ O), SO4 (δ S and δ O), and HCO3 + CO3 (δ C and δ O) were measured to understand the sources of salts to TAM soils.

Chapter 3, “Relative terrestrial exposure ages inferred from meteoric 10Be and

- NO3 concentrations in soils along the Shackleton Glacier, Antarctica” estimates the exposure ages of soils along the length of the glacier, from near the Polar Plateau to the

10 - RIS. The coupling of meteoric Be and NO3 allows for a greater number of estimated ages (compared to only using meteoric 10Be), which help constrain the advance and retreat of the Shackleton Glacier during glacial and interglacial periods.

Lastly, Chapter 4, “Geochemical zones and environmental gradients for soils from the Central Transantarctic Mountains, Antarctica” describes geochemical trends related to geography and geomorphology across the Shackleton Glacier region. Particular attention is given to total water-soluble salt concentrations, water-soluble N:P molar ratios, and

- - ClO3 and ClO4 concentrations all of which are known to influence biodiversity. The trends are then coupled with machine learning models to predict geochemical gradients based on geography and soil moisture.

A variety of geochemical techniques were used to improve our understanding of water-soluble ion geochemistry and glacial history, and to aid in our understanding of how abiotic factors may influence biogeography and ecological habitat suitability.

Figure 1.1. Map of the Transantarctic Mountains

Map of Antarctica showing the Transantarctic Mountains. The Shackleton Glacier, located in the Queen Maud Mountains, is shown in the black box. Figure borrowed and modified from United States Geologic Survey (USGS): https://lima.usgs.gov/documents/LIMA_overview_map.pdf

Figure 1.2. Shackleton Glacier surface

Image of the Shackleton Glacier flowing through exposed peaks of the Transantarctic Mountains. Photo credit to M. Diaz, 2017.

Figure 1.3. Most and least visited terrestrial environments in Antarctica

Human footprint (HFP) scores were given to exposed terrestrial surfaces across the Antarctic continent to represent soils which have been most and least visited and modified by humans. Specifically, HFP was considered to be the compaction of soils due to human activity. Lower scores indicate the least relatively modified environments. The black box indicates the location of Shackleton Glacier. Figure borrowed and modified from Pertierra et al. (2017).

Figure 1.4. Recent glacial and interglacial periods from EPICA Dome C ice core

EPICA Dome C ice core record showing eight recent glacial and interglacial cycles in Antarctica compared to other paleoclimatic records. Box (a) represents the Milankovich cycles, (b) is the δD composition of EPICA Dome C and Vostok ice cores, (c) is the δD composition of the ocean, and (d) is the concentration of dust in the Dome C ice core. The vertical black box indicates the most recent glacial maximum. Figure borrowed and modified from Augustin et al. (2004).

Figure 1.5. Modeled ice sheet height during the Last Glacial Maximum

Modeled ice sheet height (A) during the Last Glacial Maximum (LGM) and changes in ice sheet height between LGM and modern ice levels (B). Black boxes indicate the location of the Shackleton Glacier. Figure borrowed and modified from Golledge et al. (2012).

• Nutrient ratios (C:N:P) • Surface exposure age

Figure 1.6. Antarctic nematode habitat suitability model

Habitat suitability for soil nematodes in the McMurdo Dry Valleys. Figure borrowed and modified from Freckman and Virginia (1998).

Figure 1.7. Map of the Shackleton Glacier located in the central Transantarctic Mountains

Map of the Shackleton Glacier located in the central Transantarctic Mountains. Soils from the ice-free areas on either side of the glacier were sampled, analyzed, and interpreted in this work. Map is courtesy of the Polar Geospatial Center (PGC).

Figure 1.8. Aerial photo of the Shackleton Glacier Camp (85° S, 176° W)

Aerial photo of the Shackleton Glacier Camp (85° S, 176° W), taken during the 2017- 2018 austral summer. Photo credit to M. Diaz, 2017.

Chapter 2. Stable isotopes of nitrate, sulfate, and carbonate in soils from the Transantarctic Mountains, Antarctica: A record of atmospheric deposition and chemical weathering

Accepted in Frontiers in Earth Science: Novel Isotope Systems and Biogeochemical Cycling During Cryospheric Weathering in Polar Environments with coauthors: Jianghanyang Li, Greg Michalski, Thomas H. Darrah, Byron J. Adams, Diana H. Wall, Ian D. Hogg, Noah Fierer, Susan A. Welch, Christopher B. Gardner, W. Berry Lyons

2.1. Abstract

Soils in ice-free areas in Antarctica are recognized for their high salt

concentrations and persistent arid conditions. While previous studies have investigated

the distribution of salts and potential sources in the McMurdo Dry Valleys, logistical

constraints have limited our investigation and understanding of salt dynamics within the

Transantarctic Mountains. We focused on the Shackleton Glacier (85° S, 176° W), a

major outlet glacier of the East Antarctic Ice Sheet located in the Central Transantarctic

Mountains (CTAM), and collected surface soil samples from 10 ice-free areas.

- 2- Concentrations of water-soluble nitrate (NO3 ) and sulfate (SO4 ) ranged from <0.2 to

~150 µmol g-1 and <0.02 to ~450 µmol g-1, respectively. In general, salt concentrations

increased with distance inland and with elevation. However, concentrations also

- increased with distance from current glacial ice position. NO3 is derived primarily from

2- the dissolution of soda niter (NaNO3), while SO4 is likely from a variety of salts,

including gypsum or anhydrite (CaSO4 ∙ 2H2O or CaSO4) and thenardite or mirabilite

(Na2SO4 or Na2SO4 ∙ 10H2O).

To understand the source and formation of these salts, we measured the stable

- 2- isotopes of dissolved water-soluble NO3 and SO4 , and soil carbonate (HCO3 + CO3)

15 17 δ N-NO3 values ranged from -47.8 to 20.4‰ and, while all Δ O-NO3 values are

34 18 positive, they ranged from 15.7 to 45.9‰. δ S-SO4 and δ O-SO4 values ranged from

12.5 and 17.9‰ and -14.5 to -7.1‰, respectively. Total inorganic carbon isotopes in bulk soil samples ranged from 0.2 to 8.5‰ for δ13C and -38.8 to -9.6‰ for δ18O. A simple

- mixing model indicates that NO3 is primarily derived from the stratosphere (30-100%) and troposphere (0-70%), and is possibly altered by post-deposition photolysis or

2- volatilization in soils. SO4 is primarily derived from secondary atmospheric sulfate

(SAS) by the oxidation of reduced sulfur gases and compounds in the atmosphere by

H2O2, carbonyl sulfide (COS), and ozone. Calcite and perhaps nahcolite (NaHCO3) are formed through both slow and rapid freezing and/or the evaporation/sublimation of HCO3

+ CO3-rich fluids. Our results indicate that the origins of salts from ice-free areas within the CTAM represent a complex interplay of atmospheric deposition, chemical weathering, and post-depositional processes related to glacial history and persistent arid conditions. These findings have important implications for the use of these salts in deciphering past climate and atmospheric conditions, biological habitat suitability, glacial history, and can possibly aid in our future collective understanding of salt dynamics on

Mars.

2.2. Introduction

Ice-free areas within the Transantarctic Mountains (TAM) have been of scientific interest throughout the 20th and 21st centuries due in part to their unique polar desert soil environments. They are characterized by average annual temperatures below freezing, low amounts of precipitation, and low biomass. Throughout the mid to late Cenozoic, much of the currently exposed areas along the TAM were re-worked by the advance and retreat of the East Antarctic Ice sheet (EAIS), but some soils are believed to have remained primarily ice-free for possibly millions of years (Mayewski and Goldthwait,

1985; Anderson et al., 2002). As a result of persistent arid conditions since at least the

Miocene, salts have accumulated in some Antarctic soils (Marchant and Denton, 1996).

Early geochemical work in the McMurdo Dry Valleys (MDV) (77° S, 162° E), the largest ice-free area in Antarctica, showed that the soil environments in Antarctica are among the most saline systems on Earth (Jones and Faure, 1967; Keys and Williams,

1981). The binary salts, which are primarily nitrate-, sulfate-, chloride- and carbonate- bearing, have been used for determining relative chronology, and have important implications for habitat suitability and hence soil biodiversity (Claridge and Campbell,

1977; Keys and Williams, 1981; Magalhães et al., 2012; Bockheim and McLeod, 2013;

Sun et al., 2015; Lyons et al., 2016). Antarctic ice-free environments are often utilized as

Martian analogues, and salt formation processes in Antarctica may aid in our understanding of salt sources and the current and past availability of water on Mars

(Wynn-Williams and Edwards, 2000; Vaniman et al., 2004; Bishop et al., 2015).

By interpreting the relationship between the types of salts in the soils, the pH, and the distribution of calcite crusts, Claridge and Campbell (1977) and Keys and Williams

(1981) proposed that the majority of MDV salts were derived from marine sources, however, in-situ chemical weathering and deposition of oxidized atmospheric compounds

- - are also important. While Cl salts are generally derived from marine aerosols, and HCO3

- 2- salts from lacustrine deposits and chemical weathering, the origins of NO3 and SO4 salts are more complex (Claridge and Campbell, 1977; Nezat et al., 2001; Bisson et al.,

2015). Additionally, when liquid water is present, the dissolution of salts and ion exchange in soils can alter the original salt geochemistry (Toner and Sletten, 2013; Toner et al., 2013).

- 2- The measurement of stable isotopes of NO3 , and SO4 has greatly improved our understanding of the sources and transport of these salts in Antarctica. Potential sources

- of NO3 include deposition from polar stratospheric clouds, tropospheric oxidation of

HNO3 emitted from ice, nitrification and denitrification of nitrogen species by organisms, and oxidation of oceanic organic matter (Savarino et al., 2007; Frey et al., 2009a;

2- Campbell et al., 2013; Erbland et al., 2015), while potential sources of SO4 include pyrite weathering, marine biogenic sulfate, sea-salt sulfate, and S from volcanic eruptions

(Legrand and Delmas, 1984; Patris et al., 2000; Rech et al., 2003; Shaheen et al., 2013).

17 - 2- Additional work has used O isotopes to attribute NO3 , and SO4 salt abundances to the deposition of atmospheric oxidized compounds, particularly in old, high elevation, and hyper-arid environments in Antarctic ice-free areas (Bao et al., 2000; Michalski et al.,

2005; Bao and Marchant, 2006). 19

- 2- The majority of isotopic measurements of NO3 and SO4 in Antarctic terrestrial systems have been made on soils from the MDV (e.g. Bao et al., 2000; Bao and

Marchant, 2006; Jackson et al., 2016; Michalski et al., 2005; Nakai et al., 1975). Few studies have investigated the geochemistry, distribution, and source of salts from the

Central Transantarctic Mountains (CTAM), which are believed to contain some of the most saline soils on Earth (Sun et al., 2015; Lyons et al., 2016). We measured the

- 2- concentrations and isotopic composition of NO3 , SO4 , and HCO3 + CO3 in soil samples collected from the Shackleton Glacier region, located in the CTAM, to identify potential salt sources. We show that salt composition varies throughout the region, likely related to differences in the availability of water. Atmospheric deposition is the primary source of

- - both NO3 and SO4 , while carbonate minerals are formed from the freezing and evaporation/sublimation of water. These data provide insights into the processes that lead to salt formation and accumulation in CTAM soils.

2.3. Methods

2.3.1. Study site

During the 2017-2018 austral summer, a multi-disciplinary field camp was established at the Shackleton Glacier (~84.5°S), a major outlet glacier of the EAIS in the

CTAM. The Shackleton Glacier flows between several exposed peaks of the Queen

Maud Mountains, which are the basis of this study (Fig. 2.1). Though climate data for the region are sparse, winter temperatures are well below freezing and summer months are closer to 0℃ (LaPrade, 1984). Elevations of the ice-free areas range from ~150 m.a.s.l.

towards the Ross Ice Shelf to >3,500 m.a.s.l. further inland. The soils in this study were collected between ~300 m.a.s.l. and 2,100 m.a.s.l. (Table 2.1). The geologic basement consists of gneiss, schist, slate, and quartzite formed from sedimentary and igneous strata which were intruded by granitoid batholiths in the Ross Orogeny. Devonian to Triassic rocks of the Beacon Supergroup overlie the basement, which have been cut by dolerite/basaltic sills (Elliot and Fanning, 2008). Near the Ross Ice Shelf, the exposed surfaces are primarily comprised of metamorphic and igneous rocks, while the Beacon

Supergroup and Ferrar Dolerite are more abundant towards the Polar Plateau along with sediments from the Sirius Group. These rocks serve as primary sources of weathering products for salt formation (e.g., carbonates and cations).

2.3.2. Sample collection

The top 5 cm of soil was collected from 11 locations along the Shackleton Glacier

(total of 27 samples) using a clean plastic scoop, stored in Whirlpak™ bags, and shipped at -20℃ to The Ohio State University (Fig. 2.1). We attempted to collect three samples in transects perpendicular to the Shackleton Glacier or local tributary/alpine glaciers at each of the 11 locations. One sample was collected near the glacier, one near our estimate of the glacier’s trim line during the Last Glacial Maximum (LGM), and the third further inland to represent long-term exposure. The soil ages are not known, but samples from the southern portion of the region, such as Roberts Massif, are likely at least 4 Ma due to the presence of Sirius Group sediments (Hambrey et al., 2003). The sample locations represent the variable ambient summer temperatures, elevations, rock types and

landscape features characteristic of the Shackleton Glacier region, and they include soils from low elevation sites near the Ross Ice Shelf to high elevation sites near the Polar

Plateau.

GPS coordinates and elevation were recorded in the field and used to estimate the aerial distance to the Ross Ice Shelf (“distance to coast”) and the distance to the nearest glacier (Table 2.1). In the latter measurement, the term “glacier” was used to represent any glacier, including the Shackleton Glacier, tributary glaciers, alpine glaciers, etc.

While this distance does not account for topography, it can be used as an estimate of potential modern and past hydrologic influence and impact on salt formation and mobility.

2.3.3. Water-soluble leaches

The soil samples were leached at a 1:5 soil to water ratio for 24 hrs., following procedures previously described (Nkem et al., 2006; Diaz et al., 2018). The leachate was filtered through 0.4 µm Nucleopore membrane filters using a polyether sulfone (PES) filter funnel that was thoroughly cleaned with deionized (DI) water between samples. The leachate was stored in the dark at +4℃ until sample analysis. Filter blanks were collected and analyzed to account for any possible contamination from the filtration and storage process.

2.3.4. Major ions

- 2- Concentrations of water-soluble Cl and SO4 were measured using a Dionex

ICS-2100 ion chromatograph and an AS-DV automated sampler, as originally described 22

by Welch et al. (2010). Water-soluble cations (K+, Na+, Ca2+, Mg2+) were measured on a

PerkinElmer Optima 8300 Inductively Coupled Plasma-Optical Emission Spectrometer

(ICP-OES) at The Ohio State University Trace Element Research Laboratory (TERL).

- - Nitrate (NO3 + NO2 ) concentrations were measured on a Skalar San++ Automated Wet

Chemistry Analyzer with an SA 1050 Random Access Auto-sampler. The precision of replicated check standards and samples was better than 10% for all anions, cations and nutrients. Accuracy was better than 5% for all analytes, as determined by the NIST1643e external reference standard and the 2015 USGS interlaboratory calibration standard (M-

216).

2.3.5. Nitrogen and oxygen isotope analysis of nitrate

Aliquots of the sample leachates were analyzed for Δ17O and δ15N of nitrate at

Purdue University following procedures described by Michalski et al. (2005). Dissolved nitrate solutions were first injected into air-tight vials and the headspace was flushed with

Ar. The nitrate in the solutions were reduced to N2O using TiCl3 (Altabet et al., 2019), then the isotopic composition of N2O was analyzed on a Finnigan-Mat 251 isotope ratio mass spectrometer (IRMS). Mass independent fractionation of oxygen isotopes, calculated by Δ17O = δ17O - 0.52* δ18O (Δ17O± 1.0‰), are reported in units of per mille

(‰) with respect to Vienna Standard Mean Ocean Water (VSMOW) and nitrogen

15 isotopes (δ N± 0.3‰) are reported with respect to N2. Nitrate concentrations were high enough for isotopic analysis on 21 of 27 samples attempted. Though the relationship

between δ18O and δ17O was used in the Δ17O calculation, the absolute values of these isotopes could not be determined.

2.3.6. Sulfur and Oxygen isotope analysis of sulfate

The same samples that were analyzed for Δ17O and δ15N of nitrate were analyzed for δ18O and δ34S of sulfate at the University of Tennessee Knoxville. The leachates were acidified with HCl to pH ~2 to remove any dissolved carbonate/bicarbonate ions. Sulfate was then precipitated as BaSO4, after the addition of BaCl2 (~ 10% wt./vol). The precipitate was rinsed several times with DI water and dried at 80°C. The δ34S and δ18O values of BaSO4 were determined using a Costech Elemental Analyzer and a Thermo

Finnigan TC/EA, respectively, coupled to a Thermo Finnigan Delta Plus XL mass spectrometer at the Stable Isotope Laboratory at University of Tennessee (e.g.,

Szynkiewicz et al. 2020). Isotopic values are reported in units of ‰ with respect to

Vienna Canyon Diablo Troilite (VCDT) for δ34S and VSMOW for δ18O with analytical precision <0.4‰ based on replicate measurements. Sulfur sequential extractions for δ34S of sedimentary sulfur were performed on seven dried, bulk soils from seven locations following methods from Szynkiewicz et al. (2009). The samples were ground and treated

2- with 30 ml of 6 N HCl to measure acid-soluble SO4 . Then the samples were treated with

20 mL of 12 N HCl and 20 mL of 1 M CrCl2 ∙ 6H2O under N2 to dissolve disulfide to measure Cr-reducible sulfide.

Of the 27 initially prepared for δ34S and 18O analysis, eight samples had sulfate concentrations too low for sufficient precipitation of BaSO4. Therefore, in these samples,

δ34S was analyzed using a Nu Instruments multi-collector inductively coupled plasma mass spectrometer (MC-ICP-MS) at the US Geological Survey High Resolution ICP-MS laboratory, Denver with analytical precision <0.3‰ (Pribil et al., 2015).

2.3.7. Carbon and oxygen isotope analysis of carbonate

Between 5 and 10 grams of bulk soil from five locations (Roberts Massif, Bennett

Platform, Mt. Heekin, Taylor Nunatak, and Nilsen Peak) were dried and ground to fine powder using a ceramic mortar and pestle for carbonate isotope analysis at the Stable

Isotope Laboratory at Southern Methodist University. Total inorganic carbon (TIC) was measured by adding phosphoric acid kept at 90℃ to the sample, liberating the carbonate

13 18 as CO2. The C and O composition of the CO2 was measured using a dual-inlet

Finnigan MAT 252 mass spectrometer. δ13C and δ18O are reported in units of ‰ with respect to Peedee belemnite (PDB) with an overall analytical precision of ±0.2‰ or better.

2.3.8. Scanning Electron Microscopy

One sample from Schroeder Hill (SH3-2) was analyzed using a FEI Quanta FEG

250 Field Emission scanning electron microscope (SEM) equipped with a backscattered electron detector for imaging and a Bruker energy dispersive x-ray (EDX) detector for spot chemical analysis. The sample was allowed to air dry and was then affixed to an aluminum stub with carbon tape. The stub was coated using Au-Pd with a Denton Desk V precious metal coater before analysis by SEM.

2.4. Results

2.4.1. Major ion concentrations

The concentrations of all measured water-soluble ions are variable across the

2- sampling locations and span up to six orders of magnitude for SO4 (Table 2.1). In

2- - general, the most abundant anion is SO4 , however Cl is the more dominant species in samples closest to the Ross Ice Shelf, such as those at Mt. Speed and Mt. Wasko (Table

- 2.1, Fig. 2.2). NO3 concentrations vary with distance from the coast, but also within individual sample locations (Table 2.1). For example, concentrations at Mt. Augustana vary from ~1 to 23 µmol g-1. The most abundant cation is Ca2+ for nearly all soils, except for the Schroeder Hill samples, where Na+ is the most abundant and concentrations approached 700 µmol g-1. Additionally, the two Schroeder Hill samples furthest from the glacier have the lowest Ca: Mg molar ratios (0.52-0.30), indicating an enrichment of

Mg2+ compared to Ca2+, while most other samples are dominated by Ca2+. Concentrations of K+ range from <0.03 µmol g-1 at Thanksgiving Valley to 6.85 µmol g-1 at Schroeder

Hill (Table 2.1).

2.4.2. Trends in salt distributions

- - 2- - - 2- Molar ratios of NO3 /Cl , SO4 /Cl , and NO3 /SO4 with distance from the Ross

Ice Shelf are compared for the different locations in Fig. 2.2. Between the coast and

- - approximately halfway up the Shackleton Glacier, NO3 and Cl are in approximately equal proportions, but nitrate-bearing salts become more dominant closer to the Polar

Plateau (Fig. 2.2a). The trend for sulfate-bearing salts is similar. Near the coast, chloride 26

2- is about two orders of magnitude higher in concentration than sulfate. However, SO4 becomes the dominant species for most locations beyond 50 km inland, and concentrations increase to nearly four orders of magnitude higher than Cl- (Fig. 2.2b).

- 2- The molar ratio of NO3 /SO4 with distance from the coast exhibits an inverse trend compared to the species normalized to Cl- (Fig. 2.2c), where the ratio is highest near the

2- - coast. As observed with the SO4 /Cl ratio, approximately 50 km inland, sulfate becomes

2- dominant. The relative enrichment of SO4 increases further away from the coast and closer to the Polar Plateau. In general, both nitrate and sulfate have a positive relationship with distance from the Ross Ice Shelf. These results show that, contrary to trends

- observed in the MDV and the Beardmore Glacier region (83°4’ S, 171°0’ E) where NO3 was the dominant salt for inland and high elevation locations, sulfate is instead the most abundant in the Shackleton Glacier region (Keys and Williams, 1981; Lyons et al., 2016).

15 17 - 2.4.3. δ N and Δ O of NO3

Values of δ15N and Δ17O are widely variable within and between the different locations. δ15N values range from -47.8 to 20.4‰ and while all Δ17O values are positive, they range from 15.7 to 45.9‰ (Table 2.2). The highest Δ17O and δ15N values were at

- Mt. Augustana and Thanksgiving Valley, respectively. The isotopic composition of NO3 does not appear directly related to elevation and distance from the coast, though there is a slight (R2 = 0.20, p-value = 0.05) positive relationship between δ15N and distance from the glacier (Fig. 2.3a – 2.3c).

2.4.4. δ34S and δ18O in sulfate

2- 34 Despite a wide range of SO4 concentrations, δ S values are well constrained

2- between 12.5 and 15.8‰, even for the low SO4 concentration samples that were analyzed on the MC-ICP-MS (Table 2.2). δ18O values are slightly more variable, though all values are negative and range from -14.5 to -6.8‰. The highest δ34S value is from Mt.

Heekin and the most negative δ18O value is from Schroeder Hill. Values of δ34S do not vary significantly with geography, however δ18O exhibits negative trends with elevation and distance from the glacier (Fig. 2.3d – 2.3f).

2.4.5. Inorganic δ13C and δ18O of carbonate

Although dissolved inorganic carbon species were not directly measured for the soil extracts in this work, carbonate and bicarbonate minerals have been identified throughout the MDV, and therefore, these minerals are assumed to also be present in

CTAM soils (Bisson et al., 2015; Lyons et al., 2020). The amount of carbonate in the nine bulk soil samples analyzed ranges from 0.07% at Roberts Massif near the Polar

Plateau to 2.5% at Taylor Nunatak further North (Table 2.3). δ13C values are positive for all samples, ranging from 0.2 to 8.5‰, with the exception of Bennett Platform, which has a value of -13.0‰. All δ18O values are negative and range from -38.8 to -9.6‰. One sample from Taylor Nunatak (TN1-6), noted in Table 2.3, yielded NO gas which froze out of the system and interfered with the δ18O analysis.

2.5. Discussion

These Shackleton Glacier region data represent the highest southern latitude δ15N

17 - 34 18 2- 13 18 2- and Δ O of NO3 , δ S and δ O of SO4 , and δ C and δ O of CO3 measurements made on soils and soil leaches. We evaluate water-soluble ion concentrations and compare the isotopic compositions to potential source reservoirs to understand the types of salts, sources of salts and possible post-deposition alteration in remote, hyper-arid

Antarctic terrestrial environments.

2.5.1. Water-soluble salt compositions

Molar ratios of water-soluble ions and SEM images suggest that a variety of salts exist within the soils of the Shackleton Glacier region. Salt dissolution diagrams indicate that the major nitrate salt is Na(K)NO3, though some samples, such as those from the high elevation and distant locations of Roberts Massif and Schroeder Hill, have Na+ + K+ concentrations that are higher than the 1:1 dissolution line (Fig. 2.4a). These samples

+ + likely have some Na (K ) associated with HCO3 (forming nahcolite, trona, thermonatrite and/or sodium bicarbonate), as observed in MDV and Beardmore Glacier region (Bisson et al., 2015; Sun et al., 2015), or possibly bloedite (Na2Mg(SO4)2∙4H2O) in addition to

Na(K)-NO3, which was observed in the SEM images of Schroeder Hill (Fig. 2.5).

There appear to be a range of possible sulfate salts across the region and within individual samples. Anhydrite and/or gypsum (CaSO4 or CaSO4 ∙ 2H2O) have been previously identified in MDV soils (Keys and Williams, 1981; Bisson et al., 2015) and some of the Shackleton samples plot on the salt dissolution line, consistent with the

dissolution of Ca-SO4 salts. Mirabilite (Na2SO4·10H2O) and thenardite (Na2SO4), however, have also been identified in soils and aeolian material in the MDV (Keys and

+ 2+ Williams, 1981; Bisson et al., 2015; Diaz et al., 2018). High Na and SO4 concentrations which are outside the stoichiometric lines for gypsum/anhydrite are likely due to the dissolution of these salts (Fig. 2.5c – 2.5d). The Schroeder Hill SEM images

2- and EDX spot analysis show that SO4 from this location is likely from the dissolution of gypsum or anhydrite, epsomite (MgSO4 ∙ 7H2O), thenardite or mirabilite, and/or glauberite (Na2Ca(SO4)2). Mg-SO4 salts are also suggested to be abundant in Martian soils and may reflect the water content potential of the soils (Clark and Van Hart, 1981;

Vaniman et al., 2004). We also identified an unusually abundant Na-Mg-SO4 salt, possibly bloedite (Na2Mg(SO4)2 ∙ 4H2O), which, along with the other salts observed at

Schroeder Hill, was previously described at Roberts Massif (Claridge and Campbell,

1968) (Fig. 2.5). We did not observe any HCO3 and CO3 salts in the Schroeder Hill SEM images. The variability in the salt concentrations and compositions is likely due to the heterogeneous lithology of the Shackleton Glacier region, and differences in salt solubilities. While sulfate salts, such as gypsum do not readily solubilize, even with multiple wetting events, nitrate salts, such as soda niter (NaNO3), are highly soluble and only form in hyper-arid soils (Toner et al., 2013). The presence of NaNO3 and Na2SO4 salts in the high elevation and inland samples indicates that these soils likely have had prolonged arid conditions.

- 2.5.2. Stratospheric and photochemical processes as sources of NO3

15 17 - - The δ N and Δ O composition of NO3 in our samples suggests that NO3 is derived primarily from the atmosphere, with a component derived from photolysis in

- snowpack or another post-deposition alteration process. Atmospheric NO3 has a distinct

15 isotopic signature, with a δ N value of 0‰ if derived from N2, or ~ -6 to 7‰, if derived from multiple N species, and Δ17O values >15‰ (Moore, 1977; Michalski et al., 2003).

15 - However, upon deposition to the surface, δ N values of NO3 can be altered by photolysis (the breakdown of molecules due to intense and prolonged UV radiation) and volatilization in the absence of biologic activity, which could cause δ15N to either increase or decrease depending on HNO3 equilibrium between the aqueous solution and

15 - vapor (Walters and Michalski, 2015). The range of our δ N values suggests that the NO3 is not simply from oxidized N2.

All of the Shackleton Glacier region samples have Δ17O values > 15‰, indicating

17 - an atmospheric source (Fig. 2.6). Positive Δ O values are an indicator of NO3 derived from ozone and ozone-derived oxygen in the atmosphere, which has a high non-mass

17 17 - dependent O enrichment. The O signal is preserved in NO3 and is believed to only be altered by denitrification (Reich and Bao, 2018). However, because biological denitrification is thought to be a minor process in Antarctic soils (Cary et al., 2010), the

Δ17O compositions are likely minimally altered.

- Isotopic variations of N and O in NO3 have been previously measured in exposed sediments from the MDV and the Beardmore Glacier region to elucidate the source of

- NO3 to these systems. We compared Shackleton Glacier region samples to these data, and while our isotopic values are not as well constrained, they generally plot near the

- MDV and Beardmore samples (Fig. 2.6). These variations are also independent of NO3 concentration (Fig. 2.6b). In the MDV, δ15N values ranged from -9.5 to -26.2‰ and Δ17O ranged from 28.9 to 32.7‰ (Michalski et al., 2005; Jackson et al., 2015, 2016). Further

- south, isotopic compositions of NO3 along the Beardmore Glacier ranged from 1.8 to

8.8‰ and 28.4 to 33.5‰ for δ15N and Δ17O, respectively (Lyons et al., 2016). Between the two locations, Δ17O values are identical, but δ15N values are not similar. The δ15N and

17 - Δ O range for the Beardmore overlaps with measured values of atmospheric NO3

(Moore, 1977; Michalski et al., 2003) (Fig. 2.6), and Lyons et al. (2016) suggested that

- approximately 50% of the NO3 was produced in the troposphere and 50% in the

17 - stratosphere based on the high Δ O values. In other words, the NO3 in Beardmore

Glacier region soils, as in the Shackleton Glacier region, is entirely atmospheric in origin and has preserved the atmosphere isotopic signature.

- 2.5.2.1. Post-depositional alteration and snowpack emission of NO3 in Antarctica

Though the Shackleton Glacier samples are likely initially derived from the

15 15 - atmosphere, the δ N values differ from the δ N range of atmospheric NO3 (Fig. 2.4).

- While not measured in this study, our data suggest that post-deposition alteration of NO3 likely occurs in CTAM soils, potentially due to photolysis or local oxidation of N species

(either modern or ancient). Previous studies have used both direct measurements and

- theoretical models to argue that the isotopic composition of NO3 at the Antarctic surface

can be affected by post-depositional fractionation processes, particularly re-emission of

NOx (NO + NO2) from snowpack due to photolysis and evaporation of HNO3 (Savarino et al., 2007; Frey et al., 2009a; Morin et al., 2009). Photolysis has been previously documented in glacial environments in both Greenland and Antarctica where snow accumulation rates are low (Honrath et al., 1999; Jones et al., 2001), and in

- photochemical experiments using snow and NO3 which show direct production of NOx when exposed to sunlight (Honrath et al., 2000). Savarino et al. (2007) estimated the emission flux of NOy (NO + NO2 + HNO3 + HONO + 2 x (N2O5), etc.) from oxidized

7 -1 NOx species to be ΔN = 1.2 x 10 kg yr , which is similar to the flux from polar stratospheric clouds at ΔN = 6.3 x 107 kg yr-1 (Muscari et al., 2003). In other words,

- photolysis and NOx emission is an important source of NO3 to the Antarctic N cycle.

Spatial and temporal variations in the δ15N composition of Antarctic snow are similar to the variability in the Shackleton Glacier soils, reflecting both stratospheric and

- tropospheric production of NO3 . Savarino et al. (2007) found that the composition of

- NO3 in coastal Antarctica snowpack is dependent on the season, where winter is

- 15 dominated by deposition of NO3 from polar stratospheric clouds (δ N ≈ 19‰) and the summer and late-spring composition is influenced by snow reemissions of NOx and

15 HNO3 from further inland (δ N ≈ -34‰). Frey et al. (2009) measured the spatial

- 15 17 - distribution of NO3 concentrations and δ N and Δ O values of NO3 between the

Antarctic coast and the interior. They found that δ15N values were highly positive in the interior, with values >200‰, and highly negative on the coast, with values as low as -

15 - 15‰. Frey et al. (2009) and Erbland et al. (2013) show that the δ N-NO3 of snowpack 33

in the interior of the continent is positive. They argue that subsequent photolysis and

- 15 evaporation cause gaseous loss of that NO3 as NOx. This process results in enriched δ N

15 in the remaining snow, and a lighter δ N of NOx released to the atmosphere, which is

- later re-deposited elsewhere on the continent, including soils, as HNO3 .

It should be noted that no studies have directly investigated or measured post-

15 17 - depositional fractionation of δ N and Δ O in NO3 in Antarctic soils. Jackson et al.

- (2016) argued that the isotopic signature of NO3 in MDV soils is preserved and more resistant to post-depositional alteration, likely due to acid neutralization by soil carbonate

- minerals and limited light penetration into soils. These authors assumed that NO3 which was deposited directly on soil surfaces from the atmosphere was not influenced by volatilization, photolysis, or water exchange. Instead, they suggested that the soil δ15N

- values of NO3 were affected by post-depositional alteration when overlain by ephemeral snow and near glaciers due to photolysis in the snowpack, as supported by decreasing

δ15N values further from the glacier. We do not observe a trend of decreasing δ15N values further from snow and ice in our samples. However, considering the range of δ15N values and positive Δ17O values along the Shackleton Glacier, we hypothesize that our samples

- contain a mixture of NO3 produced in the stratosphere (sedimentation from polar stratospheric clouds, oxidation of NOx by ozone) and troposphere (oxidation of HNO3 by ozone, snowpack remission, and long range transport of gases and aerosols). Both atmospheric layers could be affected by evaporation and photolysis upon deposition on soils.

- 2.5.2.2. Estimating the atmospheric contribution of NO3 to Shackleton Glacier region soils

- Production, transport and alteration of NO3 in soils requires further investigation

- to effectively determine the relative fraction of NO3 derived from both stratospheric and tropospheric sources. However, we estimate the fluxes from the two reservoirs following conceptual and theoretical models from Savarino et al. (2007), Frey et al. (2009), and

- Erbland et al. (2015). Our data suggest that for inland locations in Antarctica, NO3 is deposited from the stratosphere onto the surface of the EAIS and soils of the TAM, and initially maintains a stratospheric signal (typically δ15N near 0‰ and δ17O > 15‰).

Intense UV radiation induces photolysis and mobilization of HNO3 plus other reduced N species, especially in snow, which are later re-oxidized by tropospheric ozone and re- introduced to the surface through wet and dry deposition. This mechanism is further supported by results from Jackson et al. (2016), who found that δ15N values in soils from the MDV were similar to values from aerosols near Dumont d’Urville in the coastal

Antarctic (Saravino et al., 2007).

15 15 15 훿 푁푠표푖푙 = 푓푠푡푟푎푡 ∙ 훿 푁푠푡푟푎푡 + 푓푡푟표푝(푒푚푖푡) ∙ 훿 푁푡푟표푝(푒푚푖푡) ± 푓푝표푠푡−푑푒푝 ∙

15 훿 푁푝표푠푡−푑푒푝 (2.1)

푓푠푡푟푎푡 + 푓푡푟표푝(푒푚푖푡) = 1 (2.2)

A simple mixing model using Eq. 1 and 2 can be solved to determine the relative fractions of the different atmospheric sources to the Shackleton soils. In Eq. 1, the δ15N

- 15 composition of NO3 is derived from the fractions (f) of δ N from the stratosphere (strat), photolytic emission from snowpack to the troposphere (trop(emit)), and post-depositional

- processes (post-dep). Post-depositional alteration of NO3 is still poorly understood in soils, but over prolonged periods of exposure, we anticipate that the modern influence on isotopic composition would be minimal. Therefore, we simplify the Eq. 1 and assume that stratospheric deposition and emission to the troposphere (followed by redeposition)

- 15 are the primary sources of NO3 in Eq. 2. We use end-member values of δ N ≈ 19‰ for deposition from polar stratospheric clouds to represent the stratospheric deposition and

15 - δ N ≈ -34‰ for NO3 species liberated by photolysis to represent tropospheric deposition (Savarino et al., 2007). Solving these equations, we estimate between 30% and

- 100% of NO3 is from the stratosphere and up to 70% is from the troposphere, with the exception of one sample from Nilsen Peak which appears entirely derived from the troposphere (Table 2.5). Slightly negative fraction values for Nilsen Peak and

Thanksgiving Valley are likely due to minor variations in the source isotopic compositions. However, Mt. Speed has an anomalously low δ15N value of -47.8‰, which cannot be explained by our simple model.

- 2.5.3. Primary and secondary atmospheric, and chemical weathering derived SO4

2- 2.5.3.1. Sulfide weathering as a source of SO4

34 18 2- The δ S and δ O isotopic values of SO4 indicate that sulfide weathering is

2- likely a minor source of SO4 to the Shackleton Glacier region compared to other sources, such as atmospheric deposition. Though the δ18O values are similar between the sulfide end-member and the Shackleton soil leaches, the Shackleton δ34S values are greater (Fig. 2.7). This is probably due to the low abundance of sulfide minerals in the local lithology since the δ34S signature is preserved during sulfide weathering (Balci et al., 2007).

In sedimentary rocks, sulfide is almost exclusively found as the mineral pyrite

(FeS2). Pyrite has been observed and characterized in the metasandstone of both Bowers

Terrane (Molar Formation and Pyrite Pass) and Robertson Bay Terrane near the MDV.

Further South, till found on Mt. Sirius in the Beardmore Glacier region contained detrital pyrite, which is likely the major source of pyrite for the TAM (Hagen et al., 1990). Tills of the Sirius Group can be found throughout much of the TAM at high and low elevations, but along the Shackleton Glacier the till was most abundant at Roberts Massif and Bennett Platform, with smaller outcrops observed at Schroeder Hill (Hambrey et al.,

2003). Pyrite-bearing tills have been identified in these regions, but their distributions and isotopic compositions are variable (Holser and Kaplan, 1966; Balci et al., 2007; Pisapia et

2- al., 2007; Bao, 2015). The SO4 isotopic composition of the Shackleton soils is not reflective of a predominately pyrite source.

Sulfur sequential extractions were performed on seven samples representing the range of elevations, local lithology, and glacial histories found along the Shackleton

Glacier to investigate pyrite weathering as a potential source of S. Percentages of acid- soluble sulfate and Cr-reducible sulfide (i.e. pyrite) were generally low for nearly all samples with less than 0.5% and 0.003% (~150 to 1 µmol-S g-1), respectively (Table 2.4).

Concentrations of acid-soluble sulfate were of sufficient mass for δ34S analysis only for the high elevation and further inland locations of Roberts Massif, Mt. Augustana and

Schroeder Hill. Interestingly, the acid-soluble δ34S values are only 0.2 to 0.3‰ higher than the water-soluble δ34S values, which is within our analytical error. We suggest that this extractable phase may be primarily from gypsum/anhydrite dissolution in acid and

2- therefore is of a common source with the water-soluble SO4 since the acid extraction solubilizes both the water and acid soluble constituents. For the Cr-reducible phase, S concentrations were sufficient for δ34S analysis for two samples, Mt. Heekin and Nilsen

Peak. These δ34S values are more negative than both the water-soluble and acid-soluble values at -2.3 and 12.1‰. As a comparison, Sirius Group tills had δ34S values ranging from -1.4 to +3.1 ‰, representing an isotopic composition similar to elemental sulfur

(S0). Nevertheless, it is well-known that sedimentary sulfides are isotopically variable

(Hagen et al., 1990).

Our data show that some Shackleton Glacier region soils contain sulfide (likely as

2- pyrite), however, sulfide weathering is unlikely to be a major source of SO4 . The

2- -1 concentrations of water-soluble SO4 in our samples are as high as 450 µmol g , while most samples had sulfide concentrations too low for analysis (<0.001% or 0.3 µmol-S g- 38

1). Though Mt. Heekin had quantifiable sulfide, the concentration was only 0.003% (~1

µmol-S g-1), compared to 0.19% (~50 µmol-S g-1) for water-soluble S. Unless the sulfide reservoirs were at least 100x greater in the past and experienced complete oxidation, the

2- majority of our SO4 was derived from another source, likely the atmosphere as we

2- proposed previously. Additionally, the distinct trends we observed between SO4 concentrations and isotopic composition with elevation, distance from the coast, and

2- distance from the nearest glacier suggest that similar processes are controlling SO4 formation throughout the region. Finally, the δ34S values of the Cr-reducible sulfide from

Mt. Heekin and Nilsen Peak are too negative to explain the isotopic composition of the

2- water-soluble SO4 . All the available information suggest that chemical weathering of pyrite may occur in some Shackleton Glacier region soils, but it is a minor process and is overwhelmed by an atmospheric source.

2- 2.5.3.2. Atmospheric sulfate as the primary source of SO4

We used end-member values of δ34S and δ18O for non-sea salt secondary atmospheric sulfate (SAS) (δ34S = 12.0‰, δ18O = -16.0‰), sea salt sulfate (SS) (δ34S =

22 ‰, δ18O = 10‰), and terrestrial sulfate from sulfides (TS) (δ34S = 5 ‰, δ34S = -20‰) reported by Bao and Marchant (2006) to estimate the contributions of each source to the

Shackleton Glacier region soils. We solved a three-component mixing model (Eq. 3-5)

2- for the fractions of SAS, SS, and TS comprising the observed SO4 isotopic composition.

2- With the exception of one sample from Mt. Heekin, the SO4 in our samples appears predominately derived from an SAS source, followed by SS, and lastly TS (Table 2.6). In

particular, the higher elevation and furthest inland locations, such as Schroeder Hill and

Roberts Massif, have the highest contributions from SAS (>70%). These results are similar to those from high and inland locations in the MDV (Bao and Marchant, 2006).

Though the isotopic composition of most samples can be explained by a combination of the three end members, our simple model was not sufficient for two Thanksgiving Valley samples and three additional samples from Mt. Heekin, Taylor Nunatak, and Schroeder

Hill, probably due to unaccounted variability in the values for the SAS and TS end- members. As stated in section 4.3.1, the sulfide isotopic composition in terrestrial systems is highly variable, but the least constrained end-member is likely SAS.

34 34 34 34 훿 푆푠표푖푙 = 푓푆퐴푆 ∙ 훿 푆푆퐴푆 + 푓푆푆 ∙ 훿 푆푆푆 + 푓푇푆 ∙ 훿 푆푇푆 (2.3)

18 18 18 18 훿 푂푠표푖푙 = 푓푆퐴푆 ∙ 훿 푂푆퐴푆 + 푓푆푆 ∙ 훿 푂푆푆 + 푓푇푆 ∙ 훿 푂푇푆 (2.4)

푓푆퐴푆 + 푓푆푆 + 푓푇푆 = 1 (2.5)

SAS can have a large range of δ34S, δ18O, and Δ17O values due to differences in the initial source of S and the chemical composition of the oxidizing compounds. Sulfur gases in the atmosphere (SO2) are derived from volcanic emissions, DMS oxidation from the ocean, and anthropogenic emissions. The latter is thought to comprise the least important source for Antarctica. SO2 can be oxidized by both ozone and H2O2 to form

SAS in the troposphere and stratosphere, where the oxygenic isotopic transfer is one

2- oxygen (0.25) and two oxygen (0.5) of the total four oxygen atoms in SO4 for ozone and

17 2- H2O2, respectively, which produces a positive Δ O anomaly in SO4 and a wide range of

δ 18O values (Savarino et al., 2000; Uemura et al., 2010; Bao, 2015). Additionally, SAS can be produced in the stratosphere by photolysis of carbonyl sulfide (COS), the most 40

abundant sulfur gas in the atmosphere, and by SO2 oxidation by OH radicals, which also produce positive Δ17O anomalies (Kunasek et al., 2010; Brühl et al., 2012). Though we

17 2- could not determine the Δ O composition of the Shackleton Glacier region SO4 , we suspect Δ17O would be positive and similar to the MDV and Beardmore Glacier region

(Bao et al., 2000; Bao and Marchant, 2006; Sun et al., 2015). Future measurements of

17 2- Δ O in SO4 would provide additional evidence for SAS accumulation in CTAM soils.

2.5.3.3. Accumulation of secondary atmospheric sulfate (SAS) and wetting history

34 2- The relatively small variability in δ S values indicates that the SO4 in the

Shackleton Glacier region is derived from a common, large-scale source, such as the

2- atmosphere. Additionally, when compared to the concentrations of SO4 in the water leaches, δ34S does not vary systematically indicating that the variability is not due to differences in source, but instead from varying accumulation periods (Fig. 2.7b).

Though exposure ages have yet to be determined for these areas in the Shackleton

Glacier region, modeling studies have shown that the height of the Shackleton Glacier was probably higher than current levels during the LGM (MacKintosh et al., 2011;

Golledge et al., 2013a), and likely inundated much of the currently ice-free areas near the

Ross Ice Shelf. While these surfaces were inundated, some soils closer to the Polar

Plateau may have been ice-free and would have accumulated salts from the atmospheric deposition of SAS. When the EAIS retreated in the late Pleistocene/early Holocene, the recently exposed soils could begin accumulating salts again. The small variations in δ34S

values likely reflect isotopic changes of SAS through time. Such isotopic variation can occur with changes in volcanic activity, DMS and/or MSA production, and changes in the concentrations of ozone, OH, COS, and H2O2 in the atmosphere, as reflected by the wide-range of values for Antarctic background sources in Fig. 2.7a (Legrand et al., 1991;

Bao, 2015). The variability in δ18O values is possibly due to the removal of 18O during atmospheric transport, changes in temperature, changes in the ocean isotopic composition during glacial and interglacial periods, and/or differences in the relative abundance of oxidizing atmospheric compounds. However, without the ability to decipher the

2- difference between contemporary and paleo SO4 deposits, these mechanisms remain speculative.

2.5.4. Cryogenic carbonate mineral formation and isotope equilibrium

Pedogenic carbonates in Antarctic soils are thought to be formed by authigenesis in the presence of liquid water. It is assumed that Ca2+ ions for carbonate formation are derived from the weathering of Ca-rich aluminosilicate minerals, the dissolution of primary calcite within the soils, and/or calcium associated with aeolian dust (Lyons et al.,

2020). In solution, carbonate minerals are precipitated during dissolved Ca-HCO3/CO3 saturation when the ion activity product is greater than the solubility product. In polar region soils, this typically occurs during evaporation/sublimation or cryoconcentration due to freezing of soil solutions or films (Courty et al., 1994; Vogt and Corte, 1996;

Burgener et al., 2018).

The isotopic composition of HCO3 + CO3 in the Shackleton Glacier region bulk soil samples suggests that the carbonate was originally formed by cryogenic processes, such as rapid freezing and evaporation/sublimation, with possible kinetic isotope effects

(KIE) (Fig. 2.8). Previous studies have shown that the formation of authigenic calcite deposits is controlled by dissolved CO2 concentrations and carbonate alkalinity of Ca-

HCO3 solutions (Nezat et al., 2001; Neumann et al., 2004; Lacelle et al., 2006; Lacelle,

2007; Burgener et al., 2018). The δ13C and δ18O isotopic composition of carbonate minerals is dependent on the isotopic composition and temperature of the formation fluid when in equilibrium with both the fluid and atmosphere (Lacelle, 2007). Further, the δ18O isotopic composition of the fluid is influenced by evaporation/sublimation, which depletes the fluid of the lighter oxygen isotope, and freezing, which incorporates the heavier isotope in ice (Jouzel and Souchez, 1982). However, during rapid dehydration, freezing, and carbonate dissolution, KIE can result in temperature-independent fractionation and isotopically variable carbonate species (Clark and Lauriol, 1992;

Skidmore et al., 2004; Burgener et al., 2018).

Previous studies have measured the isotopic composition of soil carbonate minerals from the MDV and have elucidated the formation mechanisms for cryogenic carbonates (see Lacelle (2007)). In summary, isotopic values of soil carbonate in Taylor and Victoria Valleys in the MDV ranged from 6.73 to 11.02‰ for δ13C and -8.13 to -

20.34‰ for δ18O (VPDB) (Burgener et al., 2018; Lyons et al., 2020). Nakai et al. (1975) measured δ13C and δ18O of carbonate coatings on rocks in the Lake Vanda Basin, MDV, and their δ13C values ranged from 1.5 to 17.6‰ while their δ18O ranged from -9.2 to - 43

31.2‰ (VPDB). Lacelle (2007) argued that the Lake Vanda basin carbonates were cryogenic in origin, forming from bicarbonate dehydration and subsequent CO2 degassing in isotopic disequilibrium. Disequilibrium during rapid evaporation/sublimation or freezing results in more positive δ13C and δ18O values relative to equilibrium carbonate formation (Clark and Lauriol, 1992; Lacelle et al., 2007). Using a clumped isotope method, Burgener et al. (2018) arrived at similar conclusions regarding disequilibrium during carbonate formation. The authors suggested that negative Δ47

(notation from clumped isotopes of mass 47) with positive δ18O anomalies, and positive

δ13C values with respect to equilibrium were consistent with cryogenic calcite formation

13 and KIE from CO2 degassing during bicarbonate dehydration. Additionally, δ C values from Taylor Valley carbonates, which were sampled as mineral coatings on rocks, were near 7.4‰ indicating an atmospheric origin of CO2 (Lyons et al., 2020), and were similar to the values reported by Burgener et al. (2018).

The Shackleton samples are generally within the range of δ13C for cryogenic carbonate in equilibrium with the atmosphere (Fig. 2.8) (Lacelle et al., 2006; Lacelle,

2007). Since we collected surface samples (up to 5 cm at depth), the carbonates were formed under conditions allowing for rapid exchange of CO2. However, some samples have lower δ18O values, possibly due to KIE, and rapid freezing and evaporation/sublimation. As stated previously, the formation of carbonate minerals in soils from rapid evaporation/sublimation of glacial meltwater results in a relatively heavier δ18O signature when compared to the ice isotopic composition (Lacelle et al.,

2006; Lyons et al., 2020). While the isotopic composition of ice in the Shackleton Glacier 44

region is unknown, due to its distance inland, we expect δ18O values ~ -45‰ (Mayewski et al., 1990; Gooseff et al., 2006). Evaporation/sublimation carbonate formation from this water may explain the relatively more positive δ18O values in the Shackleton soils compared to glacial ice. Most of our data can be explained by these mechanisms, but one sample from Bennett Platform and a second from Mt. Heekin have highly negative δ18O values, and the Bennett Platform sample is the only sample we measured with a negative

δ13C value (Fig. 2.8). These outliers demonstrate the need for more geochemical data from CTAM ice-free areas to definitively elucidate carbonate formation and kinetics in ice-free Antarctic environments.

2.6. Conclusions

Ice-free areas from the Shackleton Glacier region, Antarctica represent polar desert environments that have been modified throughout the Cenozoic, which is reflected in the variable salt geochemistry. Along a transect moving inland and up in elevation along the Shackleton Glacier towards the Polar Plateau, water-soluble salt concentrations increased, and the dominant salt species also changed. Near the Ross Ice Shelf, Cl- was

- 2- - the dominant salt, while NO3 and SO4 were more abundant further inland. High NO3

2- and SO4 concentrations are likely associated with soda niter (NaNO3), gypsum or anhydrite (CaSO4 or CaSO4 ∙ 2H2O), epsomite (MgSO4 ∙ 7H2O), thenardite or mirabilite

(Na2SO4 or Na2SO4 ∙ 10H2O) and glauberite (Na2Ca(SO4)2). We also identified abundant

Na-Mg-SO4 salts at Schroeder Hill, potentially bloedite.

15 17 - - The δ N and Δ O isotopic composition of NO3 indicated that NO3 is primarily derived from the atmosphere, with varying contributions from the troposphere (0-70%) 45

and stratosphere (30-100%). Neither δ15N or Δ17O exhibited recordable trends with elevation, distance from the coast of the Ross Ice Shelf, or distance from the glacier. We

- argue that post-depositional alteration of NO3 , potentially due to photolysis or volatilization, likely occurs in CTAM soils and possibly explains the variability in the

- NO3 isotopic composition. However, the occurrence and degree of soil photolysis of

- NO3 is unknown and requires further investigation.

2- Results from a three-component mixing model suggested that SO4 in Shackleton

Glacier region soils was predominately deposited as secondary atmospheric sulfate (SAS) and derived from the oxidation of SO2, H2S, and/or dimethyl sulfide by H2O2, COS, and

2- ozone in the atmosphere. While there is evidence to suggest that some SO4 was produced by the weathering of pyrite and other sulfide minerals, the atmospheric source was likely much more important, especially in soils which have been exposed for prolonged periods at higher elevations and near the Polar Plateau.

2- - While SO4 and NO3 were primarily derived from atmospheric deposition, carbonate minerals were formed at the surface as cryogenic carbonate. Based on the δ13C and δ18O values of soil total inorganic carbon (TIC), we conclude that both equilibrium and disequilibrium occur through slow and rapid evaporation/sublimation or freezing of fluids. Disequilibrium between the fluid and the precipitated carbonate resulted in the negative δ18O values observed due to bicarbonate dehydration.

- 2- Our analysis and interpretation of the isotopic composition of NO3 , SO4 , and

HCO3 + CO3 show that atmospheric deposition and chemical weathering at the soil

- 2- surface are important for salt formation in Antarctica. While NO3 and SO4 are both 46

oxyanions and thought to maintain their isotopic composition post-formation, post- depositional processes, such as volatilization and photolysis, may alter both N and O in

- 2- NO3 , while SO4 appears less affected by these processes. As a result, the isotopic

- - 2- composition of NO3 can potentially be used to constrain NO3 recycling in soils, SO4 can be used as an indicator of past atmospheric oxidation processes, and carbonate can be used to understand current and past availability of water. We suggest that similar processes likely occur(ed) for other hyper-arid environments in the CTAM and Mars.

Table 2.1. Salt isotope sample geographic information and concentrations of water-soluble ions in soil leaches

Sample geographic information and concentrations of water-soluble ions in soil leaches determined by ion chromatography and nutrient analysis in µmol g-1. Values are corrected for a 1:5 soil to water leach ratio. Samples that were below the analytical detection limit are listed as b.d.l. (Diaz et al., 2018).

- - - 2- Sample Feature name Latitude Longitude Elevation Distance Distance F Cl NO3 SO4 name from from coast glacier

m km m µmol g-1 µmol g-1 µmol g-1 µmol g-1

AV2-1 Mt. Augustana -85.1706 -174.1338 1410 72 388 0.35 4.53 18.5 78.1

AV2-5 Mt. Augustana -85.1691 -174.1372 1388 72 226 0.48 9.86 23.3 46.6

AV2-8 Mt. Augustana -85.1676 -174.1393 1378 72 61 0.18 0.50 1.01 0.55

BP2-1 Bennett -85.2121 -177.3576 1410 82 1007 0.30 5.26 4.13 16.2 Platform

BP2-5 Bennett -85.2072 -177.3887 1294 82 396 0.28 0.19 10.6 42.5 Platform

BP2-8 Bennett -85.2024 -177.3907 1222 82 27 1.13 0.51 0.22 0.67 Platform

MF2-1 Mt. Franke -84.6236 -176.7353 480 9 1098 0.07 0.13 b.d.l. b.d.l.

MF2-4 Mt. Franke -84.6237 -176.7252 424 9 450 0.05 0.29 b.d.l. b.d.l.

Continued

Table 2.1 continued

MH3-1 Mt. Heekin -85.0332 -177.3292 1200 63 533 0.77 25.1 27.4 142.0

MH3-5 Mt. Heekin -85.0311 -177.2489 1080 63 300 0.69 30.7 21.8 59.4

MH3-8 Mt. Heekin -85.0324 -177.2265 1045 63 502 0.23 0.81 0.36 0.26

MSP4-2 Mt. Speed -84.4657 -177.1357 n.d. 0 11 0.18 0.20 0.17 0.02

MSP4-4 Mt. Speed -84.4647 -177.1685 308 0 4 0.10 0.20 0.18 0.004

MW 4-3 Mt. Wasko n.d. n.d. 350 ~10 ~1 0.11 0.16 0.17 0.002

NP3-4 Nilsen Peak -84.5344 -175.4166 673 0 146 0.06 0.35 0.32 b.d.l.

49 RM2-1 Roberts -85.4879 -177.1844 1776 120 882 1.63 3.93 23.0 415.5

Massif

RM2-5 Roberts -85.4868 -177.1639 1754 120 706 0.57 0.10 5.66 35.6 Massif

RM2-8 Roberts -85.4857 -177.1549 1747 120 564 1.98 0.13 23.41 138.7 Massif

SH3-2 Schroeder Hill -85.3597 -175.0693 2137 94 900 0.48 0.23 149.5 218.7

SH3-5 Schroeder Hill -85.3588 -175.1198 2092 94 1214 1.23 0.39 147.5 443.6

SH3-8 Schroeder Hill -85.3569 -175.1621 2057 94 1381 0.37 0.85 93.0 126.2

Continued

Table 2.1 continued

TGV2-1 Thanksgiving -84.919 -177.0603 1107 45 1758 0.19 1.20 0.85 0.81 Valley

TGV2-5 Thanksgiving -84.9145 -176.9688 1082 45 311 0.54 49.0 15.9 34.9 Valley

TGV2-8 Thanksgiving -84.9145 -176.8860 912 45 1701 0.19 0.16 0.15 b.d.l. Valley

TN2-1 Taylor -84.9238 -176.0988 1030 45 40 0.19 5.59 1.77 0.66 Nunatak

TN2-5 Taylor -84.9264 -176.1060 1056 45 298 0.26 113.2 54.1 13.5 Nunatak

TN2-8 Taylor -84.9266 -176.1108 1070 45 350 0.20 4.74 1.40 10.3 Nunatak

Continued

Table 2.1 continued

Na+ Mg2+ K+ Ca2+ Sr2+

µmol g-1 µmol g-1 µmol g-1 µmol g-1 µmol g-1

18.9 11.3 0.53 52.2 0.07

20.6 20.1 1.06 52.1 0.04

2.15 0.26 0.09 0.24 b.d.l.

17.4 3.44 0.13 8.77 0.01

24.0 8.76 0.31 37.1 0.02

51 10.5 b.d.l. 0.09 0.10 b.d.l.

0.23 0.08 0.07 0.10 b.d.l.

0.06 0.04 0.05 0.04 b.d.l.

22.7 8.84 1.47 71.5 0.06

37.8 14.0 0.60 48.1 0.04

1.71 0.17 0.12 0.38 b.d.l.

0.48 0.08 0.18 2.48 0.001

0.04 0.03 0.08 0.80 b.d.l.

Continued

Table 2.1 continued

0.13 0.06 0.07 0.17 b.d.l.

0.31 0.10 0.13 0.44 b.d.l.

0.41 0.03 b.d.l. 0.16 b.d.l.

35.6 3.29 0.52 13.8 0.02

91.2 7.19 1.53 29.7 0.06

166.9 128.7 1.43 66.7 0.16

737.6 237.9 6.85 70.6 0.53

52 85.0 39.2 3.18 67.0 0.38

3.43 0.13 0.07 0.13 b.d.l.

21.7 20.8 0.93 30.8 0.07

0.63 0.02 0.03 0.02 b.d.l.

7.86 0.36 0.14 0.74 b.d.l.

65.9 19.3 0.94 33.9 0.08

6.84 1.72 0.31 8.04 0.01

15 17 - 34 18 2- Table 2.2. δ N and Δ O of NO3 and δ S and δ O of SO4

15 17 - 34 18 2- δ N and Δ O of NO3 and δ S and δ O of SO4 . Samples that were below the analytical detection limit are listed as b.d.l. (Michalski et al., 2005; Szynkiewicz et al., 2- 34 18 2009). Samples with low SO4 concentrations were only analyzed for δ S and not δ O (represented as n.d.).

Sample Δ17O δ15N δ34S δ18O

VSMOW Air VCDT VSMOW

AV2-1 25.9 -18.9 14.3 -9.7

AV2-5 28.0 -17.6 14.3 -9.3

AV2-8 45.9 -9.8 14.7 -7.7

BP2-1 26.9 -8.9 13.7 -9.4

BP2-5 31.9 -4.2 14.9 n.d.

BP2-8 28.5 -8.9 13.9 -8.3

MF2-1 b.d.l. b.d.l. 15.5 n.d.

MF2-4 b.d.l. b.d.l. 13.2 n.d.

MH3-1 36.0 8.8 14.2 -7.9

MH3-5 15.7 -1.2 14.1 -6.8

MH3-8 b.d.l. b.d.l. 14.0 -12.3

MSP4-2 24.9 -47.8 12.5 n.d.

MSP4-4 b.d.l. b.d.l. 13.5 n.d.

MW4-3 b.d.l. b.d.l. 15.8 n.d.

NP3-4 b.d.l. -35.1 15.0 -7.1

RM2-1 30.8 -6.1 13.4 -10.1

RM2-5 19.0 -8.3 13.8 -9.4

RM2-8 40.8 -5.0 14.1 -10.4

Continued

Table 2.2 continued

SH3-2 11.2 -3.7 13.0 -14.5

SH3-5 22.4 -12.0 13.1 -11.7

SH3-8 31.9 -6.4 13.7 -10.1

TGV2-1 19.9 20.4 12.9 -14.5

TGV2-5 n.d. 0.9 15.7 -9.2

TGV2-8 40.4 3.5 14.5 n.d.

TN2-1 18.1 9.7 15.3 -7.2

TN2-5 b.d.l. b.d.l. 14.7 -8.2

TN2-8 19.0 -4.9 15.0 -8.5

Table 2.3. δ13C and δ18O of total inorganic carbon (TIC) in bulk soil samples

δ13C and δ18O of total inorganic carbon (TIC) in bulk soil samples. TN1-6 produced NO gas which interfered with the isotopic analysis for δ18O. The sample is not included Fig. 2.8.

Sample Carb. % δ13C δ18O δ18O

PDB PDB VSMOW

BP1-4 0.24 -13.03 -39.82 -10.19

BP2-8 0.35 0.24 -20.53 9.70

MH3-10 0.38 7.13 -35.64 -5.87

NP4-6 0.11 1.40 -21.11 9.10

RM1-1 0.04 8.45 -9.55 21.01

RM3-7 0.07 4.17 -15.02 15.37

TN1-6 1.6 5.58 0.94* 31.83*

TN1-9 1.6 1.34 -15.80 14.57

TN3-5 2.5 4.21 -14.13 16.29

TN3-3 0.84 3.08 -13.93 16.51

Table 2.4. δ15 Two-component mixing model

Two-component mixing model to determine relative contributions of different sources of - 15 NO3 using Eq. 2.2-2.3. End-member values of δ N from stratospheric clouds (strat) (19‰) and tropospheric remission (trop(emit)) (-34‰) were originally reported by (Savarino et al., 2007). Negative values are indicated (*), which identify samples where the model parameters were insufficient.

Sample f(strat) f(trop(emit))

AV2-1 0.29 0.71

AV2-5 0.31 0.69

AV2-8 0.46 0.54

BP2-1 0.47 0.53

BP2-5 0.56 0.44

BP2-8 0.47 0.53

MH3-1 0.81 0.19

MH3-5 0.62 0.38

MSP4-2 -0.26* 1.26

NP3-4 -0.02* 1.02

RM2-1 0.53 0.47

RM2-5 0.49 0.51

RM2-8 0.55 0.45

SH3-2 0.57 0.43

SH3-5 0.42 0.58

SH3-8 0.52 0.48

TGV2-1 1.03 -0.03*

TGV2-5 0.66 0.34

Continued

Table 2.4 continued

TGV2-8 0.71 0.29

TN2-1 0.83 0.17

TN2-8 0.55 0.45

Table 2.5. Sulfur sequential extractions on bulk sediment

Sulfur sequential extractions on bulk sediment and δ34S values associated with each fraction. Samples that were below the analytical detection limit are listed as b.d.l.

(Szynkiewicz et al., 2009). wt. S% wt. S% δ34S δ34S

acid-solubleSO4 Cr-reducible sulfide acid-soluble SO Cr-reducible sulfide 4 VCDT VCDT TGV2-1 0.004 <0.001 b.d.l b.d.l TN2-1 <0.001 <0.001 b.d.l b.d.l MH3-5 0.001 0.003 b.d.l -2.3 AV2-1 0.268 <0.001 14.6 b.d.l NP3-4 0.001 0.002 b.d.l 12.1 RM2-8 0.432 <0.001 14.3 b.d.l SH3-2 0.87 <0.001 13.3 b.d.l

Table 2.6. δ34S and δ18O three-component mixing model

Three-component mixing model to determine relative contributions of different sources 2- 34 18 of SO4 using Eq. 2.3-2.5. End-member values of δ S and δ O for non-sea salt secondary atmospheric sulfate (SAS) (12.0 and -16.0‰), sea salt sulfate (SS) (22 and 10‰), and terrestrial sulfate from sulfides (TS) (5 and -20‰) were originally reported by Bao and Marchant (2006). Negative values are indicated (*), which identify samples where the model parameters were insufficient.

Sample f(SAS) f(SS) f(TS)

AV2-1 0.73 0.25 0.03

AV2-5 0.68 0.27 0.06

AV2-8 0.58 0.33 0.09

BP2-1 0.58 0.28 0.15

BP2-8 0.48 0.33 0.19

MH3-1 0.50 0.34 0.16

MH3-5 0.34 0.39 0.27

MH3-8 0.99 0.13 -0.11*

NP3-4 0.58 0.35 0.07

RM2-1 0.58 0.25 0.16

RM2-5 0.59 0.27 0.13

RM2-8 0.77 0.22 0.01

SH3-2 1.03 0.05 -0.08*

SH3-5 0.71 0.18 0.10

SH3-8 0.65 0.24 0.10

TGV2-1 1.02 0.05 -0.07*

TGV2-5 0.97 0.23 -0.20*

TN2-1 0.64 0.34 0.02

Continued

Table 2.6 continued

TN2-5 0.64 0.31 0.05

TN2-8 0.73 0.29 -0.02*

Figure 2.1. Study sites for salt isotope analysis

Map of Antarctica highlighting the Shackleton Glacier (SG; yellow box) in relation to the McMurdo Dry Valleys (MDV) and Beardmore Glacier (BG) (grey boxes) (a). The Shackleton Glacier is the focus of this study and flows from the Polar Plateau to the Ross Ice Shelf (b). Soil samples were collected from 10 features along the glacier: Roberts Massif (c), Schroeder Hill (d), Mt. Augustana (e), Bennett Platform (f), Mt. Heekin (g), Taylor Nunatak (h), Thanksgiving Valley (i), Mt. Wasko (j), Nilsen Peak (k), Mt. Speed (l). The symbols represent sampling locations, though no accurate GPS coordinate could be recorded from Mt. Wasko. Images of the Shackleton Glacier were acquired from the Polar Geospatial Center (PGC). 61

Figure 2.2. Molar ratios of salts with distance from coast

Molar ratios of major anion species typically associated with salts in high latitude, polar - - 2- - - 2- desert soils (NO3 /Cl , SO4 /Cl , and NO3 /SO4 ) compared to distance from the coast of the Ross Ice Shelf. The black dotted lines represent general trends in the data, while the red solid line represents equal molar proportions of the anions and the blue line represents the SO4:Cl molar ratio in seawater.

Figure 2.3. Stable isotopes of δ15N and Δ17O in nitrate and δ34S and δ18O in sulfate

Stable isotope values of δ15N and Δ17O (a-c) in nitrate and δ34S and δ18O in sulfate (d-f) with elevation, distance from the coast of the Ross Ice Shelf, and distance from the nearest glacier. Blue lines are trend lines for δ15N and δ34S, while orange lines are trend lines for Δ17O and δ18O.

- 2- Figure 2.4. Salt dissolution diagrams for the major NO3 and SO4 salts

- 2- Salt dissolution diagrams for the major NO3 (a-b) and SO4 (c-d) salts typically observed in high latitude polar desert soils. The solid black lines are dissolution lines, suggesting that samples which fall on the lines likely have salts derived from the ions represented on the x and y axes.

Figure 2.5. Scanning electron microscopy (SEM) backscatter emission (BSE) images of salt encrustations from Schroeder Hill

Scanning electron microscopy (SEM) backscatter emission (BSE) images of salt encrustations from Schroeder Hill. The chemical composition of the salts was determined by energy dispersive x-ray spectroscopy (EDX).

- Figure 2.6. Isotopic composition of NO3 for Shackleton Glacier soils compared to potential sources and other hyper-arid deserts

- 15 Isotopic composition of NO3 for Shackleton Glacier soils (a) and the variation of δ N with inverse nitrate concentrations (b). The shapes and colors representing the different sampling locations correlate with the key in Fig. 2.2. The solid black boxes represent the - isotopic composition of NO3 the mixed atmosphere, stratosphere (polar stratospheric clouds (PSC)), and troposphere (re-emission from snowpack) (Moore, 1977; Michalski et al., 2003; Savarino et al., 2007). The dashed boxes represent the Beardmore Glacier region (BG), McMurdo Dry Valleys (MDV), and Atacama Desert for comparison (Michalski et al., 2004, 2005; Jackson et al., 2015, 2016; Lyons et al., 2016). The black - 17 arrow indicates a potential NO3 formation pathway with higher δ O values from ozone. 66

2- Figure 2.7. Isotopic composition of SO4 for Shackleton Glacier soils compared to potential sources and other hyper-arid deserts

2- 34 Isotopic composition of SO4 (a) and the variation of δ S with inverse sulfate Isotopic 2- 34 composition of SO4 (a) and the variation of δ S with inverse sulfate concentrations (b). The shapes and colors representing the different sampling locations correlate with the key in Fig. 2.2. The clear circles represent MB11 samples (Asgard Tills) from the McMurdo Dry Valleys (Bao and Marchant, 2006). The solid red box shows the distribution of the 2- samples at a higher resolution. The solid and dotted black boxes represent potential SO4 2- sources and the blue star is the SO4 isotopic composition of modern seawater (Holser and Kaplan, 1966; Faure and Felder, 1981; Calhoun and Chadson, 1991; Legrand et al., 1991; Alexander et al., 2003; Pruett et al., 2004; Jonsell et al., 2005; Bao and Marchant, 2006; Baroni et al., 2008; Tostevin et al., 2014). 67

2- Figure 2.8. Isotopic composition of CO3 for Shackleton Glacier soils compared to potential sources and other hyper-arid deserts

2- 13 Isotopic composition of total inorganic carbon (CO3 ) for Shackleton Glacier soils. δ C and δ18O are reported in terms of VPDB. The shapes and colors representing the different sampling locations correlate with the key in Fig. 2.2. The solid black box represents the composition of cryogenic carbonates formed in equilibrium with the source fluid and atmosphere (slow process). The arrows represent the direction of δ18O fractionation with rapid freezing and evaporation/sublimation, and the dashed box represents samples from the McMurdo Dry Valleys (MDV) for comparison (Nakai et al., 1975; Lacelle et al., 2006; Lacelle, 2007; Burgener et al., 2018; Lyons et al., 2020).

10 - Chapter 3. Relative terrestrial exposure ages inferred from meteoric Be and NO3 concentrations in soils along the Shackleton Glacier, Antarctica

Under review in Earth Surface Dynamics with coauthors: Lee B. Corbett, Paul R. Bierman, Byron J. Adams, Diana H. Wall, Ian D. Hogg, Noah Fierer, W. Berry Lyons

3.1. Abstract

Modeling studies and field mapping show that increases in ice thickness during glacial periods were not uniform across Antarctica. Rather, outlet glaciers that flow through the Transantarctic Mountains (TAM) experienced the greatest changes in ice thickness. As a result, ice-free areas that are currently exposed may have been covered by ice at various points during the Cenozoic, thereby providing a record of past ice sheet behavior. We collected soil surface samples and depth profiles every 5 cm to refusal (up to 30 cm) from eleven ice-free areas along the Shackleton Glacier, a major outlet glacier

10 - of the East Antarctic Ice Sheet (EAIS) and measured meteoric Be and NO3 concentrations to calculate and estimate surface exposure ages. Using 10Be inventories from three locations, calculated maximum exposure ages range from 4.1 Ma at Roberts

Massif near the Polar Plateau to 0.11 Ma at Bennett Platform further north. When corrected for inheritance of 10Be from prior exposure, the ages (representing a minimum) range from 0.14 Ma at Roberts Massif to 0.04 Ma at Thanksgiving Valley. We correlate

- 10 NO3 concentrations with meteoric Be to estimate exposure ages for all locations with

- 10 - NO3 depth profiles but only surface Be data. These results indicate that NO3 69

concentrations can be used in conjunction with meteoric 10Be to help interpret EAIS dynamics over time. We show that the Shackleton Glacier has the greatest fluctuations near the Ross Ice Shelf while tributary glaciers are more stable, reflecting the sensitivity of the EAIS to climate shifts at TAM margins.

3.2. Introduction

Exposed terrestrial surfaces in Antarctica have previously been used to elucidate glacial history and assess ice sheet stability during warm periods (Denton et al., 1993;

Balco, 2011; Mackintosh et al., 2014). While Antarctica is thought to have had a permanent ice sheet since the Eocene, both the East and West Antarctic Ice Sheets (EAIS and WAIS, respectively) have fluctuated in extent and thickness throughout the Cenozoic

(Huybrechts, 1993; Barrett, 2013; DeConto and Pollard, 2016). The WAIS has been drastically reduced in size during interglacial periods and there is evidence from

ANDRILL marine sediment cores suggesting there have been numerous times over the last 11 Ma with open water in the Ross Embayment (McKay et al., 2009; Barrett, 2013;

Shakun et al., 2018). The most recent partial collapse of the WAIS was during the

Pleistocene, and the most recent total collapse was during the Pliocene (Scherer et al.,

1998; Naish et al., 2009).

The collapse of the WAIS during the Pliocene contributed ~5 m to sea level, but

Pliocene sea levels were at least 25 m higher than today, indicating additional water sources, likely from the EAIS and Greenland Ice Sheet (GIS) (Dwyer and Chandler,

2009; Pollard and DeConto, 2009). There is substantial evidence indicating that the

WAIS is susceptible to collapse due to warming (Pollard and DeConto, 2009); however,

the overall stability of the EAIS has also been questioned (Huybrechts, 1993; Wilson,

1995; Sugden, 1996; Scherer et al., 2016).

Here, we evaluated fluctuations of the EAIS during glacial and potentially interglacial periods. Outlet glaciers are among the most sensitive areas to glaciological change in Antarctica, and changes in their extents over time are recorded in nearby sedimentary deposits (Golledge et al., 2013a; Jones et al., 2015; Scherer et al., 2016;

Spector et al., 2017). We focus on the Shackleton Glacier, a major outlet glacier of the

EAIS. The Shackleton Glacier has several exposed peaks of the Transantarctic Mountains

(TAM) along the length of glacier, including at both low and high elevations. We report

10 - concentrations of meteoric Be and nitrate (NO3 ) in soils from eleven ice-free areas and use these data to calculate and estimate exposure ages. Our findings contribute to a growing body of work suggesting that some portions of the EAIS are susceptible to rapid advance and retreat.

3.3. Background

3.3.1. Stability of the EAIS

There are two competing hypotheses regarding the stability of the EAIS, though more information from various regions in Antarctica is necessary to fully refute or support either hypothesis. “Stabilists” argue that the EAIS is stable and has not fluctuated in size significantly over the last ~14 Ma (e.g., Denton et al., 1993), while “dynamicists” suggest that the EAIS is dynamic and waxes and wanes (e.g., Webb and Harwood, 1991).

Previous studies used a variety of geomorphological and exposure age dating techniques at high elevations (>1000 m) in the McMurdo Dry Valleys (MDV) to assert that the 71

Antarctic interior has maintained its aridity and cold-based glaciers since the mid-

Miocene (Sugden et al., 1993, 1995; Sugden, 1996; Lewis et al., 2008). These studies suggest major thickening of outlet glaciers but no major ice sheet retreat during the

Pliocene (Marchant et al., 1996; Golledge et al., 2013a).

Evidence for a dynamic EAIS is derived primarily from the diamictite rocks (tills) of the Sirius Group, which are found throughout the TAM and include well-documented outcrops at the Shackleton Glacier. The Sirius Group deposits are characteristic of warm and polythermal based glaciers (Hambrey et al., 2003), but their age is not known. Some of the deposits contain pieces of shrubby vegetation, suggesting that the Sirius Group formed under conditions warmer than present with trees occupying inland portions of

Antarctica (Webb et al., 1984, 1996; Webb and Harwood, 1991). Sparse marine diatoms found in the sediments were initially interpreted as evidence for formation of the Sirius

Group via glacial over riding of the TAM during the warmer Pliocene (Barrett et al.,

1992), though it is now argued that the marine diatoms were wind-derived contamination, indicating that the Sirius Group is older (Stroeven et al., 1996; Scherer et al., 2016).

Following several reviews of the stable versus dynamic EAIS debate, Barrett (2013) concluded that the EAIS maintained polar desert conditions with minimal retreat throughout the Pliocene. More recent models have suggested that portions of the EAIS, particularly outlet glaciers, were and still are susceptible to rapid retreat (DeConto and

Pollard, 2016; Scherer et al., 2016). However, the degree of EAIS sensitivity to warming is model-dependent and exposure ages/proxy data are needed to constrain model results

(Dolan et al., 2018).

3.3.2. Cosmogenic nuclide exposure age dating and meteoric 10Be systematics

10Be is a cosmogenic radionuclide with a half-life of 1.39 Ma (Nishiizumi et al.,

2007) that is produced both in the atmosphere (meteoric) and in-situ in mineral grains. In the atmosphere, N and O gases are bombarded by high energy cosmic radiation to

10 10 10 produce meteoric Be. Particle reactive BeO or Be(OH)2 is produced and removed from the atmosphere by wet and dry deposition (McHargue and Damon, 1991). At

Earth’s surface, meteoric 10Be sorbs onto clay particles and is insoluble in most natural waters of pH greater than 4 (You et al., 1989; Brown et al., 1992). Meteoric 10Be accumulation in soils is controlled by surface exposure duration, erosion, clay particle translocation, solubility, and sedimentation. Thus, meteoric 10Be can be used as a tool to understand exposure age, erosion rates, and soil residence times (see Willenbring and

Von Blanckenburg, 2009 and references within).

The measurement and use of meteoric 10Be has enabled researchers to date surfaces and features which otherwise lack sufficient coarse-grained quartz for in-situ

10Be analysis. Previous studies have measured meteoric 10Be in MDV and Victoria Land soils and sediments to calculate exposure ages and to determine the onset of the current polar desert regime (Graham et al., 2002; Schiller et al., 2009; Dickinson et al., 2012;

Valletta et al., 2015). These previous studies generally show that high elevation, northern fringe regions along the Ross Embayment have been hyper-arid since at least the

Pliocene. Meteoric 10Be data have yet to be published from the central Transantarctic

Mountains (CTAM), which represent ice sheet dynamics closer to the Polar Plateau.

Here, we used meteoric 10Be to estimate CTAM relative exposure ages, acknowledging the widespread use of in-situ exposure age dating which we later use for cross-validation. In-situ cosmogenic nuclides, such as 10Be, 26Al, 21Ne, and 3He, have been used to determine surface exposure ages at several locations across Antarctica, particularly in the MDV and other exposed surfaces in Victoria Land (e.g. (Brook et al.,

1993, 1995; Ivy-Ochs et al., 1995; Bruno et al., 1997; Strasky et al., 2009; Balco et al.,

2019). There are considerably fewer studies from the CTAM (Ackert and Kurz, 2004;

Balter et al., 2020; Bromley et al., 2010; Kaplan et al., 2017; Spector et al., 2017).

Previously reported exposure ages of CTAM tills and boulders ranged from <10 ka to

>14 Ma, and suggest that the EAIS may have maintained persistent arid conditions since as early as the Miocene. However, many previous age-date estimates were inferred from samples collected at the glacier heads and may not encompass fluctuations near the glacier terminus. Additionally, in-situ dating relies on the occurrence of coarse-grained minerals (usually quartz) in rocks and boulders, and thus is spatially limited.

3.4. Study sites

Shackleton Glacier (~84.5 to 86.4°S; ~130 km long and ~10 km wide) is a major outlet glacier of the EAIS which drains north into the Ross Embayment with other

CTAM outlet glaciers to form the Ross Ice Shelf (RIS) (Fig. 3.1). The ice flows between exposed surfaces of the Queen Maud Mountains, which range from elevations of ~150 m near the RIS to >3,500 m further inland. The basement geology of the Shackleton Glacier region is comprised of igneous and metamorphic rocks that formed from intruded and metamorphosed sedimentary and volcanic strata during the Ross Orogeny (450-520 Ma)

(Elliot and Fanning, 2008). The southern portion of the region consists of the Devonian-

Triassic Beacon Supergroup and the Jurassic Ferrar Group, while the northern portions consists of Pre-Devonian granitoids and the Early to Mid-Cambrian Taylor Group

(Paulsen et al., 2004; Elliot and Fanning, 2008). These rocks serve as primary weathering products for soil formation (Claridge and Campbell, 1968). Deposits of the Sirius Group, the center of the stable vs. dynamic EAIS debate, have been previously identified in the southern portion of the Shackleton Glacier region, particularly at Roberts Massif (Fig.

3.2) and Bennett Platform, with a small exposure at Schroeder Hill (Hambrey et al.,

2003).

The valleys and other ice-free areas within the region have been modified by the advance and retreat of the Shackleton Glacier, smaller tributary glaciers, and alpine glaciers. Similar to the Beardmore Glacier region, the Shackleton Glacier region is a polar desert, which results in high rates of salt accumulation in soils. The surface is comprised primarily of till, weathered primary bedrock, and scree, which range in size from small boulders and cobbles to sand and silt. Clays have been previously identified in all samples from Roberts Massif and are likely ubiquitous throughout the region

(Claridge and Campbell, 1968). However, the clays are a mixture of those derived from sedimentary rocks and contemporaneous weathering (Claridge and Campbell, 1968).

Thin, boulder belt moraines, characteristic of cold-based glaciers, were deposited over bedrock and tills at Roberts Massif, while large moraines were deposited at Bennett

Platform, characteristic of warm or polythermal glacial dynamics (Fig. 3.2, Balter et al.,

2020; Claridge and Campbell, 1968).

3.5. Methods

3.5.1. Sample collection

During the 2017-2018 austral summer, we visited eleven ice-free areas along the

Shackleton Glacier: Roberts Massif, Schroeder Hill, Bennett Platform, Mt. Augustana,

Mt. Heekin, Thanksgiving Valley, Taylor Nunatak, Mt. Franke, Mt. Wasko, Nilsen Peak, and Mt. Speed (Fig. 3.1). Two samples (Table 3.1) were collected at each location

(except for Nilsen Peak and Mt. Wasko, represented by only one sample) with a plastic scoop and stored in Whirl-Pak™ bags. One sample was collected furthest from the

Shackleton Glacier or other tributary glaciers (within ~2,000 m) in a transect to represent soils that were likely exposed during the Last Glacial Maximum (LGM) and previous recent glacial periods. A second sample was collected closer to the glacier (between

~1,500 and 200 m from the first sample) to represent soils likely to have been exposed by more recent ice margin retreat.

Soil pits were dug by hand at the sampling locations furthest from the glacier for

Roberts Massif, Schroeder Hill, Mt. Augustana, Bennett Platform, Mt. Heekin,

Thanksgiving Valley, and Mt. Franke. Continuous samples were collected every 5 cm until refusal (up to 30 cm) and stored frozen in Whirl-Pak™ bags. All surface (21) and depth profile (25) samples were shipped frozen to The Ohio State University and kept frozen until analyzed.

3.5.2. Analytical methods

3.5.2.1. Meteoric 10Be analysis

A total of 30 sub-samples of surface soils from all locations and depth profiles from Roberts Massif, Bennett Platform, and Thanksgiving Valley were sieved to determine the grain size at each location. The percentages of gravel (>2 mm), sand (63-

425µm), and silt (<63µm) are reported in Table A1. Since there is a strong grain size dependence of meteoric 10Be where very little 10Be is carried on coarse (>2 mm) grains

(Pavich et al., 1986), the gravel portion of the sample was not included in the meteoric

10Be analysis. The remaining soil (<2 mm) was ground to fine powder using a shatterbox.

Meteoric 10Be (Table 3.2) was extracted and purified at the NSF/UVM

Community Cosmogenic Facility following procedures originally adapted and modified from Stone (1998). First, 0.5 g of powdered soil was weighed into platinum crucibles and

0.4 g of SPEX 9Be carrier (with a concentration of 1,000 μg mL-1) was added to each sample. The samples were fluxed with a mixture of potassium hydrogen fluoride and sodium sulfate. Perchloric acid was then added to remove potassium by precipitation and later evaporated. Samples were dissolved in nitric acid and precipitated as beryllium hydroxide (Be(OH)2) gel, then packed into stainless steel cathodes for accelerator mass spectroscopy isotopic analysis at the Purdue Rare Isotope Measurement Laboratory

(PRIME Lab). Isotopic ratios were normalized to primary standard 07KNSTD with an assumed ratio of 2.85 x 10-12 (Nishiizumi et al., 2007). We corrected sample ratios with a

10Be/9Be blank ratio of 8.2 ± 1.9 x 10-15, which is the average and standard deviation of

two blanks processed alongside the samples. We subtracted the blank ratio from the sample ratios and propagated uncertainties in quadrature.

3.5.2.2. Nitrate analysis

Separate, un-sieved sub-samples of soil from all locations and depth profiles were leached at a 1:5 soil to water ratio for 24 hours, then filtered through a 0.4 µm

Nucleopore membrane filter. The leachate was analyzed on a Skalar San++ Automated

Wet Chemistry Analyzer with a SA 1050 Random Access Auto-sampler (Welch et al.,

- 2010; Lyons et al., 2016). Concentrations are reported as NO3 (Table A2) with accuracy, as determined using USGS 2015 standard, and precision better than 5% (Lyons et al.,

2016).

3.5.3. Exposure age model

We developed a mass balance using the fluxes of meteoric 10Be in and out of

Shackleton Glacier region soils to calculate the amount of time which has passed since the soil was exposed (Pavich et al., 1984, 1986). The model assumes that soils that were overlain by glacial ice in the past, and are now exposed, accumulated a lower surface concentration and inventory of 10Be than soils that were exposed throughout the glacial period (Fig. 3.3). The concentration of meteoric 10Be at the surface (N, atoms g-1) per unit of time (dt) is expressed as a function (Eq. 3.1), where the addition of 10Be is represented as the atmospheric flux to the surface (Q, atoms cm-2 yr-1) and the removal is due to radioactive decay, represented by a disintegration constant (λ, yr-1) and erosion (E, cm yr-

1) with respect to soil density (ρ, g cm-3).

푑푁 = 푄 − 휆푁 − 퐸휌푁 (3.1) 푑푡 푑푧

However, this function is highly dependent on dz, which represents an unknown value of depth into the soil column which is influenced by meteoric 10Be deposition and removal. We can account for this uncertainty and other uncertainties regarding 10Be migration in the soil column by calculating the inventory (I, atoms cm-2) of the soil (Eq.

3.2), assuming that Q has not changed systematically over the accumulation interval

(Pavich et al., 1986; Graly et al., 2010).

퐼 = ∑푁 ∙ 휌 ∙ 푑푧 (3.2)

If the inventory of meteoric 10Be in the soil profile, the concentration at the surface, and soil density are known, and published values for erosion and 10Be flux to the surface are used, we can combine Eq. (3.1) and Eq. (3.2), and solve for time (t, years)

(Eq. 3.3).

푡 = − 1 ∙ ln [1 − 퐼휆 ] (3.3) 휆 푄−퐸휌푁

Equation (3.3) provides a maximum exposure age assuming that the soil profile

10 did not have meteoric Be before it was exposed to the surface (N0 = 0). Since our exposure age dating technique relies on the number of atoms within the sediment column

10 (I), any pre-existing Be atoms in the soil (N0 ≠ 0) cause the calculated age to be an overestimate (Fig. 3.3c – 3.3d) (Graly et al., 2010). Meteoric 10Be concentrations typically decrease with depth until they reach a “background” level (Graly et al., 2010).

We can use that background value to calculate an initial inventory, also referred to as

-2 10 -1 inheritance (Ii, atoms cm ) using Eq. (3.4), where Nz is the Be concentration (atoms g ) at the bottom of the profile (z, cm), and correct the observed inventory (Eq. 3.5). 79

However, an accurate Ii can only be determined for soil profiles which have a decrease in

10Be concentrations to background levels due to the downward transport of 10Be from the surface. This may not be the case in areas of permafrost where 10Be is restricted to the active layer (Bierman et al., 2014).

퐼푖 = 푁푧 ∙ 휌 ∙ 푧 (3.4)

푡 = − 1 ∙ ln [1 − (퐼−퐼푖)휆] (3.5) 휆 푄−퐸휌푁

3.5.3.1. Model variable selection

The exposure age calculations are dependent on the selected values for the variables in Eq. (3.1-3.5). We chose a flux value (Q) of 1.3 x 105 atoms cm-2 yr-1 from

Taylor Dome (Steig et al., 1995) due to a similar climate to that of the CTAM and an absence of local meteoric 10Be flux data. While we did not calculate erosion rates, previous studies have estimated rates from rocks of 1 to 65 cm Ma-1 in Victoria Land

(Ivy-Ochs et al., 1995; Summerfield et al., 1999; Margerison et al., 2005; Strasky et al.,

2009; Morgan et al., 2010) and 5 to 35 cm Ma-1 further south in the Transantarctic

Mountains (Ackert and Kurz, 2004; Morgan et al., 2010; Balter et al., 2020). Balter et al.

(2020) determined that erosion rates for boulders at Roberts Massif which were less than

2 cm Ma-1. However, we chose a conservative value of 5 cm Ma-1 for our analysis of the

Shackleton Glacier region. Soil density (ρ) across the Shackleton Glacier region was approximately 2 g cm-3.

3.6. Results

3.6.1. Surface concentrations of meteoric 10Be and grain size

Surface concentrations of meteoric 10Be span more than an order of magnitude and range from 2.9 x 108 atoms g-1 at Mount Speed to 73 x 108 atoms g-1 at Roberts

Massif (Fig. 3.4). At individual sites where samples were collected at two locations, concentrations are typically highest for the samples furthest from the glacier, with notable exceptions at Roberts Massif and Thanksgiving Valley. In general, concentrations of meteoric 10Be increase with both distance from the coast and elevation (Fig. 3.5). There is a stronger relationship with distance from the coast (R2 = 0.48), compared to elevation

(R2 = 0.39). An exception to this trend is Bennett Platform as both surface samples from

Bennett Platform have lower concentrations than expected from the linear regression. If the samples from Bennett Platform are excluded from the linear regression, the R2 values increase to 0.67 and 0.51 for distance from the coast and elevation, respectively, with p- values < 0.001 for both regressions.

Sediment grain size is similar among the three soil profiles from Roberts Massif,

Bennett Platform, and Thanksgiving Valley; the soils are primarily comprised of sand- sized particles, with less silt-sized and smaller material (Fig. 3.6). The proportions of silt and gravel are similar at Roberts Massif, although the majority of the profile is sand- sized. Thanksgiving Valley has the least fine material, while Bennett Platform has a more even grain size distribution.

3.6.2. Calculated maximum and inheritance-corrected exposure ages

Calculated maximum meteoric 10Be exposure ages for Roberts Massif, Bennett

Platform, and Thanksgiving Valley range from 0.11 Ma at Bennett Platform to 4.1 Ma at

Roberts Massif, assuming no inheritance (Table 3.3). Bennett Platform is the only location that has exponentially decreasing 10Be concentrations with depth and appears to approach background levels towards the bottom of the 15 cm deep profile. We used the

10-15 cm 10Be concentration value to calculate the inheritance for this location. While

10Be concentrations at Roberts Massif and Thanksgiving Valley did not exponentially decrease in a similar manner, we used the lowest concentration from each of the profiles to calculate the inheritance, which is likely an overestimate. Using Eq. (3.5), the inheritance-corrected exposure ages are younger and range from 0.04 Ma at

Thanksgiving Valley to 0.14 Ma at Roberts Massif (Table 3.3). These corrected ages are minimum ages.

3.6.3. Estimated exposure ages for sites without meteoric 10Be depth profiles

- 3.6.3.1. Maximum and inheritance-corrected estimated ages using NO3 concentrations

10 - Meteoric Be and NO3 concentrations are correlated in the depth profiles from

Roberts Massif, Bennett Platform, and Thanksgiving Valley, with a strong power relationship between the two measurements (R2 = 0.66 to 0.99) (Fig. 3.7c). In addition,

10 - similar to the meteoric Be profiles, the NO3 concentrations are highest for the samples which were collected furthest from the coast and at the highest elevations (Table A2).

- 10 10 We used the relationship between NO3 and Be to estimate Be concentrations

- for all seven soil profiles (Table 3.3, Fig. 3.8). The calculated and NO3 estimated maximum exposure ages only differ by ~6-20% for Roberts Massif, Bennett Platform, and Thanksgiving Valley, which have full data sets for both parameters. The inheritance- corrected exposure ages have a difference of ~10-35% between the calculated and estimated ages. Since we could not calculate 10Be exposure ages for the profiles from

Schroeder Hill, Mt. Augustana, Mt. Heekin, and Mt. Franke, we were not able to make similar comparisons. However, we were able to compare the estimated surface 10Be

- 10 concentrations using NO3 to the measured Be concentrations. The percent differences at Schroeder Hill and Mt. Heekin are 4% and 7%, respectively, while Mt. Augustana and

Mt. Franke have higher differences of 36% and 40%, respectively (Tables 3 and 4).

3.6.3.2. Maximum estimated ages inferred using maximum meteoric 10Be concentrations

- Similar to our exposure age estimates using NO3 concentrations, we used the relationship between the maximum meteoric 10Be concentration in the soil profile and the meteoric 10Be inventory (Graly et al., 2010) to infer 10Be inventories and estimate maximum exposure ages (without a correction for inheritance) for all eleven locations

(Table 3.4, Fig. 3.8). As is the case for Roberts Massif and Thanksgiving Valley, the highest concentrations may not always be at the surface for all locations; however, the relationship is sufficiently strong to provide an estimate of the 10Be inventory and thus an age estimate (Fig. S1). Compared to the measured inventories from Roberts Massif,

Bennett Platform, and Thanksgiving Valley, the inferred inventories differ by ~3-18%.

The estimated inferred maximum exposure ages range from 0.13 Ma at Mt. Speed to >14

Ma at Roberts Massif. With the exception of Roberts Massif and Thanksgiving Valley, the oldest surfaces are those which we sampled furthest from the glacier. The sample from Roberts Massif collected closest to the glacier has an estimated exposure age that is outside the model limits (>14 Ma). The calculated maximum ages and estimated maximum ages from the inferred inventory differ by ~40% for Roberts Massif and

Thanksgiving Valley, and the estimated age is half the calculated age for Bennett

Platform (Table 3.4).

3.7. Discussion

3.7.1. Calculated and estimated exposure age validation

The Shackleton Glacier region soil profiles have the highest meteoric 10Be concentrations (~109 atoms g-1) yet measured in Earth’s polar regions (Fig. 3.7a). Though our profiles are shallower than profiles from the MDV and Victoria Land in Antarctica

(Schiller et al., 2009; Dickinson et al., 2012; Valletta et al., 2015) and Sweden and

Alaska in the Arctic (Ebert et al., 2012; Bierman et al., 2014), the soils from these previous studies reached background concentrations of 10Be within the top 40 cm, which is close to our maximum depth of 30 cm at Thanksgiving Valley. The Bennett Platform soil profile is most similar to the soil profiles from other regions in Antarctica, as they have decreasing 10Be concentrations with depth, while Thanksgiving Valley and Roberts

Massif are relatively homogenous and more similar to profiles from the Arctic. As a result, our profiles are likely sufficient for inventory and inheritance calculations.

Our calculated and estimated exposure ages are consistent with the limited in-situ exposure age data from the Shackleton Glacier region (http://antarctica.ice-d.org; Balco,

2020). From in-situ 10Be, 26Al, 3He, and 21Ne data, exposure ages on the northern flank of

Roberts Massif range from ~0.33 to 1.58 Ma (Balter et al., 2020; ICE-D), and our inheritance-corrected calculated age was 0.14 Ma, with a maximum (un-corrected) value

- of 4.09 Ma. The inheritance-corrected NO3 estimated age is 0.17 Ma. To the north, the in-situ ages from Thanksgiving Valley vary greatly from ~4.3 ka to 0.45 Ma, though most ages appear to be around 35 ka (ICE-D), which is close to our inheritance-corrected

- calculated and NO3 estimated ages of ~40 ka and ~30 ka, respectively. Closer to the Ross

Ice Shelf, the in-situ ages from Mt. Franke range from ~29 ka to 0.19 Ma. These values

- are similar to our NO3 estimated ages, which range from ~18 ka for the inheritance- corrected age to a maximum age of 0.23 Ma.

The in-situ ages are youngest closer to the glacier at nearly all locations along the

Shackleton Glacier (Balter et al., 2020; ICE-D), which is the same trend we observed for the meteoric 10Be ages. In addition, the in-situ ages and calculated and estimated ages from the Shackleton Glacier region are typically younger at lower elevations and decrease closer to the Ross Ice Shelf (Fig. 3.8). Similar patterns have been observed in the Beardmore Glacier region. Exposure ages at the head of the Beardmore Glacier at the

Meyer Desert are the oldest (up to 5.0 Ma). However, on the western side near the

Beardmore Glacier, the ages are only ~10 ka (Ackert and Kurz, 2004). To the north, ages from Cloudmaker range from ~9 ka to 15 ka near the glacier, and ~ 600 to 3 ka near the

Ross Ice Shelf at Mt. Hope (Spector et al., 2017). We argue that while the maximum

calculated and estimated exposure ages can indicate general trends in exposure ages and are useful in establishing an upper age limit, they are likely an overestimate and the inheritance-corrected (minimum) ages are more accurate, as determined by comparison to previous work.

- 3.7.2. NO3 as an efficient exposure age dating tool

- This study is not the first to attempt to use water-soluble NO3 to help understand

- glacial history, but it is the first use NO3 concentrations to directly estimate meteoric

10Be concentrations. Previous studies have argued that atmosphere-derived salt concentrations at the surface may correlate with exposure ages and wetting ages in

Antarctica (Graham et al., 2002; Schiller et al., 2009; Lyons et al., 2016; Graly et al.,

- 2018). Graly et al. (2018) showed that, in particular, water-soluble NO3 and boron exhibited the strongest relationships (R2 = 0.9 and 0.99, respectively). Lyons et al. (2016) used nitrate concentrations to estimate the amount of time since the soils were last wetted and Graham et al. (2002) attempted to calculate exposure ages using the inventory of nitrate in the soil. Graly et al. (2018) argue that boron is preferable to nitrate due to concerns related to nitrate mobility under sub-arid conditions (e.g. Frey et al., 2009;

Michalski et al., 2005), and given that uncertainties in local accumulation rates and ion

- transport can result in inaccurate ages when using NO3 alone (Graham et al., 2002;

Schiller et al., 2009). Based on the results presented here for hyper-arid CTAM ice-free regions and the concerns with boron mobility depending on whether the B species present

3- - in the soils is BO3 (borate) or H3BO3 (boric acid), we conclude that NO3 appears suitable for relative age dating and for producing age estimates. 86

10 - Through a coupled approach using both meteoric Be and NO3 concentrations, we developed a useful model for estimating soil exposure ages. We show that the percent

- differences between calculated and NO3 estimated ages are low (see Section 3.6.3.1.) and

10 - argue that the relationship between meteoric Be and NO3 can be used to expand our current exposure age database for the TAM; compared to cosmogenic radionuclide

- - analyses, NO3 analyses are rapid and cost effective. However, a model using NO3 or salts alone is likely insufficient, unless the anion accumulation rates are known (Graham

- 10 et al., 2002; Schiller et al., 2009). Though the regressions between NO3 and Be are strong (Fig. 3.7c), each of the three profiles from Roberts Massif, Bennett Platform, and

Thanksgiving Valley have different regression coefficients and slopes. In other words,

10 - the relationship between meteoric Be and NO3 is not uniform across the Shackleton

Glacier region and varies depending on the location. This is likely due to local glacial history and climate, soil development, and geography. To address these uncertainties, some 10Be data, surface samples for all locations and a few depth profiles in particular, are necessary to choose the proper regression to minimize the associated error.

10 - We tested our meteoric Be – NO3 model with data from Arena Valley (Graham et al., 2002) and found that our model is roughly applicable to other TAM ice-free areas.

10 - The power relationship between Be and NO3 throughout the profile is not as strong for the Arena Valley samples compared to Shackleton Glacier samples; there is a stronger correlation in the top 20 cm (R2 = 0.61) than the bottom 70 cm (R2 < 0.01). The estimated inventory is 7.22 x 109 atoms cm-2, while the calculated inventory is 1.3 x 1010 atoms cm-

2, and the exposure ages (without erosion and inheritance corrections) are 56 ka and 87

- ka, respectively. Though our inheritance-corrected NO3 estimated ages are validated using in-situ data from previous studies, until our estimated exposure dating technique can be tested more broadly, we interpret these ages as relative or estimated ages.

3.7.3. Implications for ice sheet dynamics

Sirius Group deposits were only observed at Roberts Massif (Fig. 3.2a) and were either deposited or exposed as the Shackleton Glacier retreated in this region. At sample site RM2-8, where soil was collected closest to the Shackleton Glacier, we documented a large diamictite that is underlain by soils estimated to be a maximum of >14 Ma in age.

While this soil age is likely an overestimate given previously published in-situ ages

(Balter et al., 2020), the Sirius Group was not observed near the relatively younger RM2-

1 soils, with an inheritance-corrected age of 0.14 Ma. We interpret these sparse data to suggest that either the tills were transported from further inland during previous glacial retreat, or that the Sirius Group formed over an extended period of time. However, considering we did not observe any diamictite on younger soils, these observations support previous studies (e.g. Barrett, 2013; Sugden et al., 1993, 1995; Sugden, 1996), which argue that, at least for the southern Shackleton Glacier region, the Sirius Group likely formed prior to the Pliocene.

Our data support models and previous studies suggesting that EAIS advance and retreat was not synchronous during the LGM and throughout the late Cenozoic (Marchant et al., 1994; Golledge et al., 2013a; DeConto and Pollard, 2016; Scherer et al., 2016).

Calculated and estimated exposure ages (including both maximum and inheritance- corrected) are youngest near the coast and greatest at the head of the Shackleton Glacier 88

(Fig. 3.8). The furthest inland sample at Mt. Franke indicates that deglaciation occurred as recently as ~0.02 Ma in the northern portion of the region, although the samples closest to the glacier are likely younger in age and may indicate that deglaciation continued into the late Pleistocene/ early Holocene (Spector et al., 2017). Deglaciation in the southern portion of the region likely occurred earlier, with the furthest inland samples from Roberts Massif, Schroeder Hill, and Bennett Platform exposed since shortly before or after the onset of the last glacial period (~0.10 Ma) (Blunier and Brook, 2001; Clark et al., 2009; Mackintosh et al., 2014). Previous data from Roberts Massif also suggests that much, if not all of this location was ice-free throughout the last glacial period (Balter et al., 2020). However, our inferred maximum estimated ages also indicate that, similar to the more northern locations, the samples collected closest to the glacier are likely younger and were more recently exposed due to ice retreat (Fig. 3.8).

Tributary glaciers in the Shackleton Glacier region appear to behave differently than the Shackleton Glacier itself. This is best demonstrated by the Bennett Platform samples, collected near the tributary Gallup Glacier. Bennett Platform is unique in being the only location we sampled with large lateral moraines and several nearby medial moraines (Fig. 3.2c). The surface concentration of meteoric 10Be is lower at Bennett

Platform than what would be expected from regression models relating concentration with elevation and distance from the coast (Fig. 3.5). The lower concentrations of 10Be, in turn, result in relatively lower calculated and estimated exposure ages (Fig. 3.8; Table

3.3). Specifically, the exposure ages suggest that glacier retreat following termination of the last glacial period was delayed at Bennett Platform.

We argue that the younger than anticipated exposure age is due to differences in glacial dynamics between tributary and major outlet glaciers. Meteoric 10Be concentrations and exposure ages at Mt. Augustana are also lower than anticipated given its distance from the coast and elevation. Similar to Bennett Platform, Mt. Augustana is along a tributary glacier, McGregor Glacier. We did not observe the same large moraines from Bennett Platform, but it is possible that McGregor Glacier and Gallup Glacier behave similarly and have a comparatively delayed response to the transition from glacial to interglacial periods. Previous work in the Royal Society Mountains found that marine and land-terminating glaciers behave asynchronously; although sea-level rise likely induced grounding line retreat in the Ross Sea following the LGM, alpine glaciers have since advanced (Higgins et al., 2000; Jackson et al., 2018). The Shackleton Glacier is marine terminating and likely susceptible to ice shelf stability and sea level rise, while the regional tributary glaciers are likely grounded on bedrock troughs and are resulting more stable with respect to changes in climate. Though the physical properties of Gallup and

McGregor Glaciers are unknown during the LGM and previous glacial periods (i.e. cold vs. polythermal, shallow vs. deep grounding), these glaciers possibly represent the dynamics of other tributary glaciers in the CTAM, which may similarly have a delayed response to climate shifts.

3.8. Conclusions

10 - We measured concentrations of meteoric Be and NO3 in soils from eleven ice- free areas along the Shackleton Glacier, Antarctica, which include the highest measured meteoric 10Be concentrations from the polar regions. Calculated maximum and

inheritance-corrected (minimum) exposure ages are well-correlated with estimated ages,

- 10 determined using NO3 concentrations and inferred Be inventories. In particular,

- 10 coupling NO3 concentrations with Be measurements represents an efficient method to attain a greater number of exposure ages in the CTAM, a region with currently sparse

- 10 data. However, while the relationship between NO3 and Be is strong in the Shackleton

Glacier region, its widespread applicability needs to be further evaluated.

Soil exposure ages are generally youngest at lower elevations and closer to the

Ross Ice Shelf, but are also younger closer to the Shackleton Glacier or other tributary glaciers. Though we could only estimate maximum inferred ages, our soil transects likely encompass the LGM transition. Inheritance-corrected calculated and estimated ages at

Roberts Massif (~1 km from the glacier) indicate that the Shackleton Glacier was likely present in its current form since at least the Pleistocene in southern portions of the region.

More northern samples indicate that towards the glacier terminus, the Shackleton Glacier is more susceptible to changes in climate and has likely retreated in the past. However, tributary glaciers likely had a delayed retreat following the LGM. These data represent a comprehensive analysis of meteoric 10Be to demonstrate the dynamic behavior of CTAM outlet glaciers at glacier termini and stability at glacier heads.

Table 3.1. Meteoric 10Be sampling locations

Geographic data of samples collected from eleven ice-free areas along the Shackleton Glacier. Distance from the coast (aerial) was measured post-collection using ArcMap 10.3 software. Samples of the format “X-1” are samples collected furthest from the glacier in the transect.

Location Sample name Latitude Longitude Elevation Distance from (m) coast (km) Mt. Augustana AV2-1 -85.1706 -174.1338 1410 72 Mt. Augustana AV2-8 -85.1676 -174.1393 1378 72 Bennett Platform BP2-1 -85.2121 -177.3576 1410 82 Bennett Platform BP2-8 -85.2024 -177.3907 1222 82 Mt. Franke MF2-1 -84.6236 -176.7353 480 9 Mt. Franke MF2-4 -84.6237 -176.7252 424 9 Mt. Heekin MH2-1 -85.0299 -177.2405 1098 63 Mt. Heekin MH2-8 -85.0528 -177.4099 1209 63 Mt. Speed MSP2-1 -84.4819 -176.5070 270 0 Mt. Speed MSP2-4 -84.4811 -176.4864 181 0 Mt. Speed MSP4-1 -84.4661 -177.1224 276 0 Mt. Wasko MW4-1 -84.5600 -176.8177 345 10 Nilsen Peak NP2-5 -84.6227 -176.7501 522 0 Roberts Massif RM2-1 -85.4879 -177.1844 1776 120 Roberts Massif RM2-8 -85.4857 -177.1549 1747 120 Schroeder Hill SH3-2 -85.3597 -175.0693 2137 94 Schroeder Hill SH3-8 -85.3569 -175.1621 2057 94 Thanksgiving TGV2-1 -84.9190 -177.0603 1107 45 Valley Thanksgiving TGV2-8 -84.9145 -176.8860 912 45 Valley Taylor Nunatak TN3-1 -84.9227 -176.1242 1097 45 Taylor Nunatak TN3-5 -84.9182 -176.1282 940 45

Table 3.2. Meteoric 10Be concentrations

Table 3.2: Concentration of meteoric 10Be in Shackleton Glacier region surface soils and depth profiles from Roberts Massif, Bennett Platform, and Thanksgiving Valley.

Sample Sample Mass AMS Uncorrected Uncorrected Background- Background- 10Be 10Be name mass of 9Be Cathode 10Be/9Be ratio 10Be/9Be corrected corrected concentration concentration (g) added Number (10-11)** ratio 10Be/9Be ratio 10Be/9Be ratio (109 atoms g-1) uncertainty (μg)* uncertainty (10-11)*** uncertainty (107 atoms g-1) (10-13)** (10-13)*** AV2-1 0.499 394.3 151135 2.201 1.143 2.201 1.143 1.162 0.604 AV2-8 0.500 400.2 151137 1.786 1.067 1.785 1.067 0.955 0.571 BP2-1, 0-5 0.499 401.2 151147 1.616 1.055 1.615 1.055 0.868 0.567 BP2-1, 5-10 0.499 399.2 151148 0.353 0.748 0.352 0.748 0.188 0.400 BP2-1, 10-15 0.496 400.2 151149 1.573 1.894 1.573 1.894 0.848 1.021

93 BP2-8 0.498 400.2 151550 0.542 0.448 0.541 0.448 0.291 0.241

MF2-1 0.505 398.2 151554 3.713 3.444 3.712 3.444 1.956 1.815 MF2-4 0.501 398.2 151555 2.448 1.395 2.447 1.396 1.300 0.741 MH2-1 0.498 399.2 151138 0.864 0.820 0.863 0.820 0.462 0.439 MH2-8 0.499 395.3 151139 0.681 0.847 0.680 0.847 0.360 0.449 MSP2-1 0.499 403.2 151556 0.539 0.464 0.538 0.464 0.291 0.250 MSP2-4 0.502 402.2 151557 0.693 0.673 0.692 0.674 0.370 0.361 MSP4-1 0.499 400.2 151566 1.112 1.117 1.111 1.117 0.596 0.598 MW4-1 0.498 400.2 151564 1.093 0.662 1.092 0.662 0.586 0.356 NP2-5 0.496 402.2 151565 2.391 1.200 2.391 1.200 1.295 0.650 RM2-1, 0-5 0.502 399.2 151558 8.541 4.116 8.541 4.116 4.538 2.187 RM2-1, 5-10 0.499 398.2 151559 8.853 8.411 8.852 8.411 4.721 4.485 RM2-1, 10- 0.500 400.2 151560 13.70 8.460 13.70 8.460 7.327 4.524 15 RM2-8 0.498 401.2 151561 10.17 15.27 10.17 15.27 5.475 8.221 Continued

Table 3.2 continued SH3-2 0.497 398.2 151551 7.191 3.129 7.190 3.129 3.850 1.675 SH3-8 0.501 398.2 151552 4.270 3.351 4.269 3.351 2.267 1.780 TGV2-1, 0-5 0.498 398.2 151140 1.860 2.431 1.859 2.431 0.993 1.299 TGV2-1, 5- 0.500 398.2 151141 1.731 1.589 1.731 1.589 0.921 0.846 10 TGV2-1, 10- 0.497 393.3 151142 1.635 1.377 1.634 1.377 0.864 0.728 15 TGV2-1, 15- 0.502 399.2 151143 1.645 1.776 1.645 1.777 0.874 0.944 20 TGV2-1, 20- 0.498 403.2 151144 1.711 0.852 1.710 0.852 0.925 0.461 25 TGV2-1, 25- 0.497 399.2 151145 2.148 2.071 2.147 2.071 1.152 1.112 30 TGV2-8 0.499 399.2 151146 2.106 2.185 2.105 2.185 1.125 1.168 TN3-1 0.500 401.2 151562 7.092 5.903 7.091 5.903 3.802 3.165 94 TN3-5 0.500 401.2 151563 3.926 5.694 3.925 5.694 2.105 3.053

*9Be was added through commercial SPEX carrier with a concentration of 1000 μg mL-1. **Isotopic analysis was conducted at PRIME Laboratory; ratios were normalized against standard 07KNSTD3110 with an assumed ratio of 2850 x 10-15 (Nishiizumi et al., 2007). Blank 10Be/9Be ratio values averaged 8.152 ± 1.884 x 10-15.

- Table 3.3. Calculated and NO3 estimated exposure ages

- Exposure ages calculated from Eq. (3.1-3.5) and estimated ages using NO3 concentration data.

Location Calculated Calculated Calculated Calculated Estimated Estimated Estimated Estimated inventory inheritance max exposure exposure age inventory inheritance max exposure age (1011 atoms (1011 atoms age (106 yrs) with (1011 atoms (1011 atoms exposure with cm-2) cm-2) inheritance cm-2)* cm-2)* age (106 inheritance (106 yrs) yrs)* (106 yrs)* Augustana - - - - 0.58 0.55 0.60 0.03 Bennett 0.13 0.06 0.11 0.07 0.14 0.06 0.12 0.07 Franke - - - - 0.27 0.25 0.23 0.02 Heekin - - - - 0.65 0.63 0.70 0.02 Roberts 1.47 1.36 4.09 0.14 1.51 1.37 4.54 0.17

95 Schroeder - - - - 1.05 0.98 1.66 0.08

Thanksgiving 0.57 0.52 0.54 0.04 0.47 0.43 0.43 0.03 - 10 *Estimations derived from power relationship between NO3 concentration and meteoric Be concentration

Table 3.4. Estimated exposure ages from maximum 10Be concentration

Estimated exposure ages using relationship between maximum 10Be concentration and inventory in Fig. A1 (Bierman et al., 2014).

Sample Estimated inventory (1011 Estimated max exposure name atoms cm-2) age (106 yrs)

AV2-1 0.38 0.35 AV2-8 0.33 0.30 BP2-1 0.31 0.27 BP2-8 0.31 0.27 MF2-1 0.21 0.17 MF2-4 0.18 0.15 MH2-1 0.59 0.62 MH2-8 0.42 0.40 MSP2-1 0.16 0.13 MSP2-4 0.18 0.15 MSP4-1 0.24 0.20 MW4-1 0.24 0.20 N2-5 0.42 0.39 RM2-1 1.24 2.65 RM2-8 1.50 >14* SH3-2 1.07 1.75 SH3-8 0.67 0.74 TGV2-1 0.34 0.31 TGV2-8 0.38 0.35 TN3-1 1.06 1.70 TN3-5 0.62 0.68 *Outside of model range

Figure 3.1. Overview of Shackleton Glacier region meteoric 10Be sampling locations

Overview map of the Shackleton Glacier region, located in the Queen Maud Mountains of the Central Transantarctic Mountains. The red circles represent our eleven sampling locations, with an emphasis on Roberts Massif (orange), Bennett Platform (green), and Thanksgiving Valley (blue), which have the most comprehensive dataset in this study. The bedrock serves as primary weathering product for soil formation (Paulsen et al., 2004; Elliot and Fanning, 2008). Base maps provided by the Polar Geospatial Center. 97

Figure 3.2. Sirius Group and glacial moraines

The Sirius Group was documented at Roberts Massif near the RM2-8 sampling location (a). Cold-based glacier moraines were observed at Roberts Massif (b) and large polythermal moraines were observed at Bennett Platform (c).

Figure 3.3. Conceptual diagram of meteoric 10Be accumulation in Shackleton Glacier soils

Conceptual diagram of meteoric 10Be accumulation in soils during glacial advance and retreat. In “ideal” conditions, 10Be accumulates in exposed soils and 10Be concentrations beneath the glacier are negligible (a). As the glacier retreats, 10Be can begin accumulating in the recently exposed soil and an inventory can be measured to calculate exposure ages. In the case where the glacier has waxed and waned numerous times and the soils already contain a non-negligible background concentration of 10Be, inventories need to be corrected for 10Be inheritance (c-d).

Figure 3.4. Spatial distribution of surface meteoric 10Be

Spatial distribution of surface meteoric 10Be concentrations in the Shackleton Glacier region. Where possible, two samples were collected at each location to represent surfaces closest to the glacier, which might have been glaciated during recent glacial periods, and samples furthest from the glacier that are likely to have been exposed during recent glacial periods. Insets of Roberts Massif (orange), Bennett Platform (green), and Thanksgiving Valley (blue) are included (color scheme consistent throughout), as these locations serve as the basis for our relative exposure age models. Base maps provided by the Polar Geospatial Center.

100

Figure 3.5. Concentration of meteoric 10Be with elevation and distance from coast

Concentration of meteoric 10Be with elevation and distance from coast. The solid black lines are linear regressions.

101

Figure 3.6. Grain size of soil pits from Roberts Massif, Bennett Platform, and Thanksgiving Valley

The grain size composition of soil profiles collected from Roberts Massif (a, orange), Bennett Platform (b, green), and Thanksgiving Valley (c, blue). The soil pits from Bennett Platform and Thanksgiving Valley are also shown with distinct soil horizons.

102

10 - Figure 3.7. Depth profiles of meteoric Be and NO3

Soil profiles of meteoric 10Be concentrations for Roberts Massif (orange), Bennett Platform (green), and Thanksgiving Valley (blue) compared to profiles from the Antarctic (Dickinson et al., 2012*; Schiller et al., 2009†; Valletta et al., 2015‡) and Arctic (Bierman et al., 2014¶; Ebert et al., 2012§) (a). The 10Be concentration profiles were also - compared to NO3 concentration profiles (b) and a power function was fit to the data (c).

103

Figure 3.8. Estimated maximum exposure age versus distance from the coast and elevation

Estimated maximum age versus distance from the coast (a) and elevation (b). The blue - triangles represent the maximum age estimates using the relationship between NO3 and 10Be, black and white triangles represent maximum age estimates using inferred 10Be inventories. Upward facing triangles are samples collected furthest from the glacier, while downward triangles are samples collected closest to the glacier. Sample RM2-8 (Roberts Massif, closest to glacier) is outside the range. Linear regression lines are - plotted for the three datasets where the solid line is for the NO3 estimate, the dashed line is the inferred estimate for samples furthest from the glacier, and the dotted line is the inferred estimate for samples closest to the glacier. The estimated maximum ages (blue) - and inheritance-corrected ages (grey) using the NO3 concentrations are overlaid on a map of the Shackleton Glacier region (c).

104

Chapter 4. Geochemical zones and environmental gradients for soils from the Central Transantarctic Mountains, Antarctica

In preparation for Antarctic Science with co-authors: Christopher B. Gardner, Susan A. Welch, W. Andrew Jackson, Byron J. Adams, Diana H. Wall, Ian D. Hogg, Noah Fierer, W. Berry Lyons

4.1. Abstract

Previous studies have established links between biodiversity and soil

geochemistry in the McMurdo Dry Valleys, Antarctica, where environmental gradients

dictate soil biodiversity. However, these gradients are not well established in the Central

Transantarctic Mountains, which are thought to represent some of the least hospitable

Antarctic soils. We analyzed 220 samples from eleven locations along the Shackleton

Glacier (~85 °S), a major outlet glacier of the East Antarctic Ice Sheet. We established

three zones of distinct geochemical gradients near the head of the glacier (upper), central

(middle), and at the glacier mouth (lower). The upper zone had the highest water-soluble

salt concentrations with total salt concentrations of >80,000 µg g-1, whereas the lower

zone had the lowest water-soluble N:P ratios, suggesting that in addition to other

parameters (such as proximity to water/ice), the lower zone might host the most favorable

ecological habitats. Given the strong dependence of geochemistry with geographic

parameters, we established multiple linear regression and random forest models to predict

soil geochemical trends given latitude, longitude, elevation, distance from the coast,

105

distance from the glacier, and soil moisture (variables which are generally remotely measurable). Confidence in our model predictions was moderately high, with R2 values

- - for total water-soluble salts, water-soluble N:P, ClO4 , and ClO3 of 0.51, 0.42, 0.40, and

0.28, respectively. These modeling results can be used to predict geochemical gradients and estimate salt concentrations for other Transantarctic Mountain soils.

4.2. Introduction

From an ecological standpoint, the least biologically diverse terrestrial systems are in extreme physical and chemical environments. The abundance and diversity of life in soils is dependent on a number of environmental parameters, including temperature, precipitation, organic matter content, and nutrient availability (Wall et al., 2012). Hot deserts are typically viewed as one of the least biologically diverse environments; however, cold deserts have even less taxa (Freckman and Virginia, 1998). Soils in

Antarctica typically serve as end-members for low habitat suitability due to their high salt concentrations, low organic carbon, low soil moisture, and low mean annual temperatures

(Courtright et al., 2001).

In the McMurdo Dry Valleys (MDV), organic matter and salt concentrations influence soil communities, where higher amounts of organic carbon, lower water-soluble

N:P ratios, and lower total dissolved solids generally result in the greatest biomass and biodiversity (Barrett et al., 2006; Magalhães et al., 2012; Caruso et al., 2019; Bottos et al., 2020). These Antarctic ecosystems are simple and are the only known soil systems where nematodes and collembola are at the top of the food chain (Freckman and Virginia,

1998). The investigation of soils in the MDV and Transantarctic Mountains (TAM )has

106

been important in understanding ecosystem structure and function in extreme terrestrial environments (e.g. Caruso et al., 2019; Collins et al., 2019; Freckman and Virginia,

1998).

In Antarctic soils, ecosystem function appears to be dependent on the availability, duration, and proximity of soils to liquid water (Barrett et al., 2006). Due to the seasonality in freezing and thawing events, liquid water acts as a pulse to the ecosystem, providing water for organisms, but also wetting surface soils and dissolving soluble salts

(Zeglin et al., 2009; Webster-Brown et al., 2010). Investigations of the salt thresholds of

Antarctic nematodes found that no organisms survived in highly saline soils (~2,600 mg

L-1) (Nkem et al., 2006). Concentrations of soluble salts exist at these concentrations or higher for high elevation and inland locations in the TAM (Bockheim, 2008; Lyons et al.,

2016). Additionally, studies on TAM soils have found that increased salt concentrations led to a decrease in ecosystem community complexity in older soils when compared to younger soils (Magalhães et al., 2012). Yet, despite these inhospitable conditions (e.g. high salt concentrations and glacial advance and retreat), some organisms are postulated to have found suitable refugia in TAM soils and have persisted in isolation for millions of years and through glacial cycles (Stevens and Hogg, 2003; Stevens et al., 2006).

It is generally accepted that habitat suitability for invertebrate species in Antarctic soils is driven by a combination of geochemical, geographic, and geomorphologic variables (Freckman and Virginia, 1998; Courtright et al., 2001; Magalhães et al., 2012;

Bottos et al., 2020). Geographic variables, such as elevation, can be easily measured with advanced mapping tools and satellite imagery; however, surface exposure ages, soil

107

geochemistry and nutrient content require extensive logistical support and resource allocation for collection and analysis. More efficient estimation tools are needed to aid in our understanding of widespread habitat suitability throughout the TAM.

In this study we have determined and evaluated geochemical patterns and gradients of water-soluble ions in soils collected from eleven locations along the

Shackleton Glacier, Central Transantarctic Mountains (CTAM). Particular attention was

- - given to total water-soluble salt concentrations, N:P ratios, and ClO4 and ClO3 concentrations, given their influence on biodiversity, as determined in previous studies

(e.g. Ball et al., 2018; Barrett et al., 2006b; Courtright et al., 2001; Dragone et al., 2020;

Nkem et al., 2006). The geochemical data were compared to geographic parameters in order to understand how the physical environment influences the observed geochemical variability. Our results show that water-soluble ion concentrations and distributions are driven largely by soil geography and surface exposure age. Finally, we implemented machine learning and statistical techniques to interpolate and predict the soil geochemistry across the region using geographic variables. Our multiple linear regression and random forest models show that latitude, longitude, elevation, distance from the coast, distance from the glacier, and soil moisture (all variables currently or soon to be remotely measurable using maps and satellites) are moderately effective at estimating trends in TAM soil geochemistry, with R2 values as high as 0.87. These data will be particularly useful for ecologists seeking to understand refugia and habitat suitability in

Antarctica and similarly harsh environments.

108

4.3. Study Sites

The Shackleton Glacier (~84.5 to 86.4°S; ~130 km long and ~10 km wide) is a S-

N trending outlet glacier of the East Antarctic Ice Sheet (EAIS) located to the west of the

Beardmore Glacier. It flows through the Queen Maud Mountains (CTAM) into the Ross

Sea (Fig. 4.1). The elevations of exposed soils range from ~150 m to >3,500 m from the coast towards the Polar Plateau. Long-term climate data are not yet available, but the

Shackleton Glacier region is a polar desert regime, similar to the Beardmore Glacier region, with average annual temperatures well below freezing and little precipitation

(LaPrade, 1984).

During the Last Glacial Maximum (LGM) and glacial periods throughout the

Pleistocene, the size and thickness of the EAIS has been suggested to be greater than current levels (Nakada and Lambeck, 1988; Talarico et al., 2012; Golledge et al., 2013b;

Wilson et al., 2018). Outlet glaciers, such as the Shackleton Glacier, may have had the greatest increases in extent, especially towards the glacier terminus (Golledge and Levy,

2011; Golledge et al., 2012). The behavior of local alpine and tributary glaciers is not well-constrained, but these glaciers are also believed to have advanced and retreated over the last two million years (Jackson et al., 2018; Diaz et al., 2020a). As a result, currently exposed soils were overlain and reworked by fluctuations of the Shackleton Glacier and other tributary and alpine glaciers in the region. Exposure ages range from the early

Holocene to the Miocene, and generally increase with distance from the coast and distance from the glacier (Balter et al., 2020; Diaz et al., 2020a).

109

The soils contain a range of water-soluble salts derived primarily from atmospheric deposition and chemical weathering (Claridge and Campbell, 1968; Diaz et al., 2020b). The major salts are typically nitrate and sulfate salts, especially at higher elevations and further inland from the coast of the Ross Sea (Diaz et al., 2020b). The solubilities of the salts vary, but nitrate salts are highly soluble and their occurrence at high elevation and inland locations suggests that those soils have persistent arid conditions.

4.4. Methods

4.4.1. Sample collection and preparation

During the 2017-2018 austral summer, 230 surface soil samples (~top 5 cm) were collected from 12 distinct ice-free areas along the Shackleton Glacier. At each location, we collected samples in transects to maximize the geochemical variability. Our transects were also designed to capture the LGM transition, with some soils exposed throughout the LGM and others exposed following glacier retreat. GPS coordinates and elevations were recorded for each sample and later used to estimate the distance from coast and distance from the glacier (Shackleton, tributary, or alpine). Once collected, the samples were stored and shipped frozen (-20 ℃) to The Ohio State University.

Before geochemical analysis, the samples were dried at 50 ℃ for at least 72 hours with the loss in mass attributed to soil moisture content. The dried soils were leached at a

1:5 soil to deionized (DI) water ratio, and the leachate was filtered through 0.4 µm

Nucleopore membrane filters (Nkem et al., 2006; Diaz et al., 2018, 2020b). Due to the low sediment to water ratio, this leaching technique only dissolves the most water-soluble 110

- - - 2- - salts (Toner et al., 2013). These include salts with ClO4 , NO3 , Cl , SO4 , ClO3 , and

2- CO3 . Blanks were generated and analyzed to account for any contamination from the leaching process. The resulting geochemical dataset interpreted in this study included 220 of the original 230 samples from 11 locations (Roberts Massif, Schroeder Hill, Mt.

Augustana, Bennett Platform, Mt. Heekin, Thanksgiving Valley, Taylor Nunatak, Mt.

Franke, Mt. Wasko, Nilsen Peak, and Mt. Speed), including a subset of 27 samples previously analyzed for S, N, O isotopes in nitrate and sulfate (Diaz et al., 2020b).

4.4.2. Analytical analysis of water-soluble anions, cations, and nutrients

The analytical techniques used here are similar to those reported by Diaz et al.

- - - 2- (2020b). To summarize, anions (F , Cl , Br , and SO4 ) were measured on a Dionex ICS-

2100 ion chromatograph, while cations (K+, Na+, Ca2+, Mg2+, and Sr2+) were measured on a PerkinElmer Optima 8300 Inductively Coupled Plasma-Optical Emission Spectrometer

- - 3- (ICP-OES), and nutrients (NO3 + NO2 , PO4 , H4SiO4, and NH3) were measured on a

Skalar San++ Automated Wet Chemistry Analyzer at The Ohio State University.

- - Perchlorate (ClO4 ) and chlorate (ClO3 ) were measured using an ion chromatograph- tandem mass spectrometry technique (IC-MS/MS) at Texas Tech University (Jackson et al., 2012, 2015). All analytes are reported as listed. Total water-soluble salt concentration was calculated as the sum of all measured cations and anions. The precision of replicated check standards and samples was typically better than 10% for all major anions, cations and nutrients, and better than 20% for perchlorate and chlorate. Accuracy was typically better than 5% for all major anions, cations, and nutrients, as determined by the

111

NIST1643e external reference standard and the 2015 USGS interlaboratory calibration standard (M-216). Accuracy better than 10% for perchlorate and chlorate, as determined by spike recoveries. Precision and accuracy for individual analytes are located in Table

B1. Detection limits for the analytes have been previous reported (Jackson et al., 2012;

Diaz et al., 2018).

4.4.3. Data interpolation and machine learning

Inverse distance weighted (IDW) interpolations were performed for Bennett

Platform, Thanksgiving Valley, and Roberts Massif using the Geostatistical Analyst tool in ArcMap 10.3. Since IDW is a deterministic method where unknown values are predicted based on proximity to known values, we chose those three sites as they had the most distinguished transects and relatively higher sample density. The interpolation parameters were constant with a power of 4, maximum neighbors of 15, minimum neighbors of 5, 4 sectors, and a variable search radius. These parameters were chosen such that they optimized for the lowest mean absolute error.

Multiple linear regressions were generated for all geochemical analytes, excluding

H4SiO4 (total of 15 dependent variables), with latitude, longitude, elevation, distance from the coast, distance from the glacier, and soil moisture as independent variables using built-in functions in R 3.6.3 (R Core Team, 2020). Random forest regression models were similarly generated using the randomForest library. The random forest model is a machine learning algorithm that uses supervised learning algorithms to predict values given input predictor variables (Breiman, 2001). Multiple decision trees are run in

112

parallel with a randomized subset of predictor variables. The aggregate result of each tree is used to generate a predicted outcome. Since each tree is generated using a random sample and random predictor variables, the random forest model is effective at minimizing overfitting and handling outliers (Breiman, 2001).

Machine learning algorithms have been widely used in a variety of disciplines from finance (Patel et al., 2015) to ecology (Prasad et al., 2006; Peters et al., 2007;

Davidson et al., 2009), for both data prediction (regression) and classification. Recently, these techniques have been used for Earth Science applications, including geologic mapping (Heung et al., 2014; Kirkwood et al., 2016), air quality monitoring (Stafoggia et al., 2019), and water contaminant tracing (Tesoriero et al., 2017). We developed a novel application of machine learning to predict concentrations and gradients of water-soluble salts in Antarctic soils, given set geographic parameters, similar to the approaches developed for stock market and real estate predictions (Antipov and Pokryshevskaya,

2012; Patel et al., 2015).

For our random forest models, any sparse missing values in Table B2 were estimated by averaging the geochemistry of the samples immediately before and after in the same transect. Missing values due to concentrations below the detection limit were input as 0. The new imputed dataset was split into a training set representing 86% of the data (n = 189, Table B3) and a testing set representing the remaining 14% (n = 31, Table

B4). The training dataset was used to generate the random forest models for each analyte.

Each of the models were run with 2000 decision trees (ntree = 2000) to minimize the mean square errors. The number of random variables used for each node split in the

113

decision trees was set to the recommended regression default of variables/3 to optimize the model randomness, which in our case was 2 (mtry = 2), following parameters described previously (Breiman, 2001). The codes developed for both the multiple linear regression and random forest models are included in Appendix A.

4.5. Results

4.5.1. Geochemistry of upper, middle, and lower glacier zones

The maximum, minimum, mean, standard deviation and coefficient of variation are reported in Table 1 for the measured geographic and geochemical data.

Concentrations of water-soluble ions span up to five orders of magnitude and are variable across the region. Elevation, distance from the coast, distance from the glacier and soil moisture are also variable and span up to three orders of magnitude. The highest elevation samples (> 2,000 m) were collected from Schroeder Hill and the greatest soil moisture content is from Mt. Wasko at 12.3%.

Shackleton Glacier region surface soils can be separated into three zones based on their water-soluble geochemistry: an upper zone near the Polar Plateau, a middle zone near the center of the glacier, and a lower zone where the glacier flows into the Ross Sea

(Figs. 4.1; 4.2). The upper zone samples are characterized by the highest total water- soluble salt concentrations (calculated as the sum of all major ions), with the highest values of greater than 80,000 µg g-1 at Schroeder Hill. The lower zone samples have the lowest total salt concentrations, with the lowest values near 10 µg g-1 at Mt. Wasko. This is generally the trend for the water-soluble N:P molar ratios as well. The lowest N:P ratios are in the lower zone soils, while the middle and upper zones have more variable 114

- - values. Concentrations of ClO4 and ClO3 generally follow similar trends as the total salts, with less distinction between middle and upper zones, though most concentrations in the lower zone are below the detection limit (Table B2).

Observed trends between the zones appear to be driven, at least partially, by geography. Regressions of total water-soluble salt concentration, water-soluble N:P ratio,

- and ClO3 concentration, with elevation, distance from the coast, and distance from the glacier are all positive (Fig. 4.2). The strongest relationships are between total salts and

- 2 elevation, and ClO3 and distance from the coast, with R values of 0.26 and 0.24, respectively, and p-values < 0.001 for both relationships (Fig. 4.2a; 4.2k). The weakest

- relationships are between ClO4 and distance from the coast and distance from the glacier, with R2 values of 0.01 (Fig. 4.2h; 4.2i). Distance from the glacier varies widely between individual zones with frequent overlaps, but there appears to be a moderate relationship with N:P ratio and total salts (Fig. 4.2c; 4.2f). Overall, total salts has the strongest

- relationships with geography and ClO4 has the weakest relationships.

Ternary diagrams highlight the specific geochemical variations within and

2- - - between the zones. The anion ternary diagram only included SO4 , NO3 , and Cl , which are the major water-soluble salts in the region (Claridge and Campbell, 1968; Diaz et al.,

2020b). Though carbonate and bicarbonate salts have been identified in both lacustrine sediments and soils in Antarctica, previously measured concentrations in the Shackleton

Glacier region were low, ranging from 0.07 to 2.5%, and bicarbonate salts were not identified in the highest elevation and furthest inland soils (Claridge and Campbell, 1968;

Perez-Lopez et al., 2007; Lyons et al., 2016). The most abundant anion for the upper zone

115

2- is SO4 , which is greater than 99% of the anion population in some Schroeder Hill and

Roberts Massif samples (Fig. 4.3). The anions are more evenly distributed in the middle

- - zone, though NO3 and Cl are the most abundant in the majority of samples. The lower

2- zone has much lower SO4 percentages than the upper zone and the dominant anion is generally Cl-. The cation distribution is very similar for all three zones (Fig. 4.3). Na+ +

K+ is the most abundant cation pair at over 90% for many upper and middle zone samples, while Ca2+ is the second most abundant. In general, Mg2+ is the least abundant cation.

4.5.2. Statistical geochemical variability

A principal component analysis (PCA) was performed in R (using factoextra and built in software libraries) to determine which geochemical variables most strongly distinguish the samples. For the PCA, the first two principal components account for over

50% of the total dataset variability at 44.2% and 11.6%, respectively. The different zones are correlated with different principal components (Fig. 4.4). The samples from the middle zone are positively correlated with PC1 and PC2. In the biplot, they plot in the

- - upper right quadrant with high concentrations of Cl , NO3 , water soluble N:P ratio, and

2+ Ca , with a minor influence from soil moisture and H4SiO4. The upper zone samples

2+ 2- 2+ + + - generally plot along PC1 and are most associated with Sr , SO4 , Mg , Na , K , F ,

- - ClO4 , and ClO3 . The samples from the lower, more coastal zone are negatively

3- correlated with PC1 and are readily distinguished by their higher PO4 concentrations.

116

Most samples from all locations plot within the 95% confidence interval ellipses, but there are two strong outliers from Schroeder Hill and Mt. Heekin.

Similar to the PCA, we performed a simple Spearman’s rank correlation for the entire dataset to visualize the statistical dependence between all variables. A goal of this study is to relate water-soluble ion concentrations to geography; therefore, we focused on latitude, longitude, distance from the coast, distance from the glacier, and soil moisture.

The strongest correlation coefficients are between Cl- and latitude, elevation and distance from the coast, and Sr2+ and soil moisture (Fig. 4.5). Most other correlations are moderate

- to weak, though there appear to be notable stronger correlations between ClO3 and latitude and distance from coast; Ca2+ and longitude, elevation, and distance from coast;

- 2- NO3 and latitude; and SO4 with distance from glacier. Longitude, elevation, and distance from coast have the most strong and moderate correlations with the geochemistry data. Outside of the geographic parameters, Na+ is highly correlated with total water-soluble salts, likely representative of the high Na+ + K+ percentages in Fig.

4.3, and Sr2+ is highly correlated with K+, likely reflecting a similar source.

4.5.3. Interpolation and machine learning model performance

The total salt concentrations of individual samples at Bennett Platform produce the strongest interpolation gradient from the glacier front to further inland compared to

Roberts Massif and Thanksgiving Valley. Bennett Platform also has the smoothest contours suggesting that the interpolation is the strongest at this location. The second strongest interpolation is Thanksgiving Valley. Contrary to Bennett Platform,

117

Thanksgiving Valley has the highest salt concentrations in the center of the valley, with lower concentrations measured to both the east and west. The lowest concentration contours are closest to the glacier for both Bennett Platform and Thanksgiving Valley, which is likely related to glacial history since the soils near the glacier are relatively younger than those further inland based on meteoric 10Be data (Diaz et al., 2020a). The interpolation from Roberts Massif does not have a distinguishable spatial trend.

The multiple linear regression and random forest models vary in their strength for the individual analytes. The highest R2 value from the linear regression is 0.55 for Sr2+,

- - while total water-soluble salts, water-soluble N:P ratio, ClO4 , and ClO3 have values of

0.37, 0.37, 0.10, and 0.33, respectively (Table 4.2). The lowest R2 value is for Cl- at 0.05.

The p-values for nearly all analytes are <<0.001, with Cl- having the only value above

0.05. The highest out-of-the-bag (OOB) explained variance values from the random

+ 2+ - 3- - forest models are for K and Sr at 62% for both analytes. Values for NO3 , PO4 , ClO4 , and N:P ratio are negative. The explained variance for total salts is 45% and the variance

- for ClO3 is 43%. We also evaluated the most important and least important variables based on node purity. The most important variable for the majority of analytes is elevation, while distance from the glacier is most important for N:P ratio and latitude for

- ClO3 (Table 4.2). The least important variable is distance from the coast for every

- analyte, except ClO3 and NH3, for which distance from the glacier is least important.

118

4.6. Discussion

4.6.1. Implications for ecological habitat suitability

By establishing geochemical zones for the Shackleton Glacier region, we can better understand the relationship between geochemistry and geography, and ultimately biogeography. As stated in the introduction, we focused particularly on total water-

- - soluble salt concentrations, water-soluble N:P ratios, and ClO4 and ClO3 concentrations.

4.6.1.1. Elevation and moisture controls on total water-soluble salt gradients

The elevational trends of total salt concentrations at the Shackleton Glacier are similar to those previous described in the TAM, where higher elevations typically correlate with higher salt concentrations (Magalhães et al., 2012; Lyons et al., 2016;

Bottos et al., 2020). Our results are also consistent with those from Scarrow et al. (2014), who found that salt concentrations typically decreased with distance from the glacier. Our total water-soluble salt interpolation maps highlight the spatial variability in Shackleton

Glacier region soils (Fig. 4.6). The most spatially variable location is Robert Massif, which does not appear to follow local elevational, latitudinal, and/or distance inland gradients. This heterogeneity is not necessarily due to current soil leaching, as the soil moisture values were not drastically different between the samples (Table B2). Though the variability in cation concentrations is likely due to weathering of tills, scree, and bedrock (Claridge and Campbell, 1968), previous work on the isotopic composition of water-soluble nitrate and sulfate, the major anions in the upper zone, suggests a common, atmospheric source (Diaz et al., 2020b).

119

We argue that the heterogeneity in the total salt concentrations at Roberts Massif

(Figs. 4.2; 4.6) is probably related to different and complex wetting history, where seasonal snow patch melt may pool in local depressions, transporting water-soluble salts from slightly higher elevations and/or from saline wet-patches (Levy et al., 2012). This is demonstrated on a larger scale at Thanksgiving Valley, a U-shaped glacially carved valley, where the higher concentrations of salts in the center of the valley are possibly due to the transport of salts from nearby higher elevation slopes during melting events.

This is further evidenced by the presence of two small ponds in the center of the valley, which likely formed from glacial melt and may have been larger in size during the recent past (Diaz et al., 2019). Similarly, streams and meltwater tracks in the MDV leach soils and carry salts into closed basin, brackish to hyper-saline lakes, where salts are cryoconcentrated over time (Lyons et al., 1998). Our results suggest that elevation and wetting history are important contributors to total salt gradients in the Shackleton Glacier region, as they influence the accumulation of salts and subsequent leaching from soils.

4.6.1.2. Influence of glacial history on water-soluble N:P ratios

Stoichiometric dependencies have been identified for Antarctic terrestrial organisms, where nutrient concentrations, in addition to soil aridity, limit ecosystem development (Nkem et al., 2006). Nitrate is primarily derived from atmospheric deposition and phosphorus is liberated from minerals by chemical weathering in the

CTAM. Therefore, many inland and higher elevation soils have accumulated high

- 3- concentrations of NO3 , resulting in a stoichiometric imbalance with soluble PO4 (Nkem et al., 2006; Barrett et al., 2007; Lyons et al., 2016; Ball et al., 2018; Diaz et al., 2020b). 120

As in the MDV, younger and coastal soils at lower elevations in the Shackleton Glacier region have the lowest water-soluble N:P ratios, driven by relatively lower concentrations

- 3- of NO3 and higher concentrations of PO4 due to an increase in moisture content and chemical weathering (Heindel et al., 2017). (Fig. 4.2; 4.4). It is unsurprising that these soils were more abundant with life; in the field, several rocks had thick lichen coverings and both collembola and mites were easily identified at Mt. Speed and Mt. Wasko (Fig.

B1). However, despite overall elevational and latitudinal gradients, some inland locations in the middle and upper zones have water-soluble N:P ratios near those from the lower zone (Fig. 4.2).

The interpolation model from Bennett Platform shows that some locations near the glacier have lower total water-soluble salt concentrations, as is the case in the MDV

(Bockheim, 2002). However, the samples near the glacier at Bennett Platform not only have lower total salt concentrations, but they also have lower N:P ratios than samples collected from further inland. This is also the case for the middle zone locations (Fig.

4.2f). We argue this is due to differences in glacial history between the locations. Our previous work shows that the soils near the glacier are younger than the soils further inland in the Shackleton Glacier region (Diaz et al., 2020a). Not only were the younger soils shielded from nitrate accumulation during glacial periods, but the recently exposed

3- rocks likely serve as fresh mineral weathering material for PO4 mobilization (Heindel et al., 2017). Recently exposed and relatively nutrient rich soils might be important refugia for invertebrates. Previous hypotheses have suggested that organisms dispersed to higher elevations during glacial periods, however abiotic gradients in the Beardmore Glacier

121

region suggest that higher elevation soils have salt concentrations that would classify them as unsuitable habitats (Stevens and Hogg, 2003; Bennett et al., 2016; Lyons et al.,

2016). If few biological community members survived glaciations, the near-glacier, relatively P-rich soils may be important in helping communities recover and restructure.

- - 4.6.1.3. High and variable ClO4 and ClO3 concentrations

- - Our ClO4 and ClO3 concentrations include the highest measured in Antarctica to date and are comparable to concentrations from the Atacama and Mojave Deserts

- (Jackson et al., 2015). The highest elevation samples (upper zone) have the highest ClO4

- and ClO3 concentrations, though a strong correlation was not observed (Fig. 4.2g; 4.2j).

- - - Similar to NO3 , ClO4 and ClO3 are derived from atmospheric deposition and because of their solubilities, their accumulations are related to wetting and glacial histories (Jackson et al., 2015, 2016). Therefore, soils which have been exposed for long periods of time and have not experienced snow or ice melt, such as those from Schroeder Hill and Roberts

- - Massif, are able to accumulate high concentrations of ClO4 and ClO3 . Interestingly, our

- ClO4 concentrations are lower than the highest recorded tolerance (1.1M NaClO4) for the extremotolerant bacteria Planococcus halocryophilus, yet a recent study shows no detectable biomass for Schroeder Hill samples (Dragone et al., 2020). Perchlorates are well established as toxic and are strong oxidizers, and it appears that the concentration of

- - ClO4 might be an additional, crucial indicator of habitat suitability. However, ClO4 concentrations are highly heterogenous across our sample locations (Fig. 4.2k – 4.2l), and

- unlike ClO3 , neither the multiple linear regression nor the random forest models were able to adequately capture the variability (Table 4.2). 122

4.6.2. Machine learning as a tool to predict soil geochemical trends

We sought to evaluate our multiple linear regression and random forest models by using a testing dataset from the Shackleton Glacier region (n = 31) plus a second dataset from the Darwin Mountains (~80°S) (n = 10) (Magalhães et al., 2012). As stated in

Section 3.3, the Shackleton Glacier region test data were not included in the random forest model generation in order to unbiasedly evaluate our models. For the Darwin dataset, distance from the glacier, distance from the coast, and elevation were determined using the reference elevation model of Antarctica (REMA), while location, soil moisture and geochemistry were retrieved from the literature (Magalhães et al., 2012). We evaluated all 15 analytes from the original models with the Shackleton dataset and due to a lack of data, only evaluated 7 analytes from the Darwin soils.

Both the multiple linear regression and random forest model outputs are moderately well-correlated for the Shackleton dataset, as determined by Pearson correlations between the measured and predicted values (Fig. 4.7a; Table 4.3). The random forest models out-perform the linear regression models for nearly every analyte,

- + - with the notable exceptions of F , Na , and NO3 , and nearly all p-values are <0.001.

Mg2+ performs the best, with R2 values of 0.79 and 0.52 for the random forest and linear regression models, respectively (Fig. 4.7a). In terms of our analytes of interest regarding habitat suitability, total salts have the strongest correlation in the random forest model (R2

2 - 2 - 2 = 0.51), followed by water-soluble N:P ratio (R = 0.42), ClO4 (R = 0.40), and ClO3 (R

= 0.28). The N:P ratio in particular performs significantly better than the linear regression model (R2 = 0.05). Mean absolute error (MAE) and root mean squared error (RMSE) 123

values indicate that the random forest models also have a smaller error compared to the multiple linear regression models (Table 4). MAE values are lower than RMSE values for both models, indicating the presence of outliers in the testing dataset. This is unsurprising as the standard deviation and coefficient of variation values for the entire dataset are relatively large for all analytes.

Similar to the model performance in the Shackleton Glacier region, the water- soluble ion predictions for the Darwin Glacier region are more strongly correlated with measured values in the random forest models when compared to the multiple linear regression models (Fig. 4.7b). In fact, the linear regression models fail for the Darwin samples and all concentration outputs are negative, which is likely due to overfitting during model generation. MAE and RSME values for both models are much higher than those for the Shackleton dataset (Table 4). On the other hand, the random forest models perform particularly well for some analytes. Though a small sample size, the R2 values for Mg2+ and K+ are 0.87, with p-values <<0.001. Total salts is moderately correlated (R2

= 0.44) and N:P ratio has an R2 value of 0.01, indicating poor model performance. It is unclear why Mg2+ and K+ performed the best, though we suspect that this is due to weathering trends of local lithology across the TAM, since chemical weathering is probably the major source of these ions.

It should be noted that the R2 values simply measure the strength of the correlations between the measured and predicted values. We performed slope tests by fitting bivariate lines using the standardized major axis (SMA) to further understand the relationship between the two values using the smatr library in R (Warton et al., 2012).

124

For this test, we specifically evaluated the null hypothesis (H0) where slope = 1, which would indicate whether an ideal direct 1:1 relationship exists between the measured and predicted values. Test statistic values (t) were used to measure the sample correlation between the residuals and fitted values (Warton et al., 2012). Test statistic values near 1 indicate that we reject the null hypothesis. In other words, higher test statistic values indicate a slope other than 1. Of the 15 analytes in the Shackleton dataset, 7 analytes have slopes near 1 for the multiple linear regression models and 6 for the random forest model, as indicated by test statistic values less than 0.5. For the Darwin samples, only one

- analyte, NO3 , has a test statistic value less than 0.5 (Fig. 4.7; Table 4.3).

These data indicate that while some analytes have high correlations between measured and predicted values, the models perform best with the Shackleton Glacier region soils. However, though the relationship may not be 1:1, the random forest models are effective at generating relative concentration data to predict geochemical gradients.

For example, similar to our data, the Darwin Glacier samples generally have greater water-soluble N:P ratios and total water-soluble salt concentrations further from the glacier and at higher elevations (Magalhães et al., 2012), a trend that is also reflected by our model results despite offset values. Additionally, corrections for the offset of the model from a slope = 1 (i.e. multiplying the model output value by the regression slope) can be made to better estimate specific concentrations, though the difference between modeled and measured values can still be up to 2x greater. Our sample size for building the multiple linear regression and random forest models was small. We anticipate that as

125

more data are collected throughout the CTAM, these data can be added to the model training dataset, expanding our prediction capabilities and increasing model reliability.

4.7. Conclusions

We established environmental and geochemical gradients in the Shackleton

Glacier region that will aid in our understanding of the abiotic properties in soils governing biodiversity. The 220 samples we analyzed from the Shackleton Glacier region represent a wide range of soil environments: those with different elevation, latitude, longitude glacial history, and geochemistry. We determined three soil zones: 1) an upper zone near the head of the glacier which is characterized by high total water-soluble salt

- - concentrations, high water-soluble N:P ratios, and high ClO4 and ClO3 concentrations,

3- 2) a lower zone with low total salt concentrations and higher PO4 concentrations, and 3) a middle zone with intermediate values. The zones help elucidate the controls of geography on soil geochemistry. In addition, our total water-soluble salt interpolations at

Roberts Massif, Bennett Platform, and Thanksgiving Valley reflect the local variability of salt concentrations and possible influences from soil age and wetting history.

Five geographic variables (latitude, longitude, elevation, distance from the coast, and distance from the glacier) and soil moisture were correlated with soil geochemistry.

We used these variables to develop multiple linear regression and random forest models.

The model results generally reflected the measured geochemical variability across the region. Test datasets from the Shackleton and Darwin Glacier regions showed that random forest models typically outperformed multiple linear regression models when correlating measured and predicted values, especially for the Darwin region. Though

126

most correlations did not exhibit a 1:1 relationship and had varying slopes, the random forest models were able to adequately predict geochemical gradients, as demonstrated by moderate to high R2 values between measured and model predicted concentrations.

127

Table 4.1. Overview of geography, soil moisture, and water-soluble ions

Overview of geography, soil moisture, and water-soluble ions from the Shackleton Glacier region. The minimum values reported are those within the detection limits. Individual sample concentrations are detailed in Table S2.

Max Min Mean STD CV Elevation (m) 2,221 150 1,125 551 48 Distance from coast (km) 120 1 55 38 68 Distance from glacier (m) 1,942 1.00 519 472 90 Soil moisture (%) 12.3 0.1 2.1 2.1 102 F- (µg g-1) 120 0.39 8.87 11.78 133 Cl- (µg g-1) 13,555 1.59 615 1,780 289

- -1 NO3 (µg g ) 38,435 0.10 1,465 3,449 235

2- -1 SO4 (µg g ) 55,266 0.08 4,385 8,078 184

3- -1 PO4 (µg kg ) 4,195 76.09 381 560 147

- -1 ClO4 (µg kg ) 75,000 0.35 985 6,019 611

- -1 ClO3 (µg kg ) 14,500 1.00 1,167 2,496 214 Ca2+ (µg g-1) 4,404 0.55 839 1,163 139 Mg2+ (µg g-1) 6,281 0.12 293 705 240 Na+ (µg g-1) 25,260 0.39 1,141 2,878 252 K+ (µg g-1) 440 0.86 28.31 51.61 182 Sr2+ (µg g-1) 46.61 0.01 8.63 10.31 119

-1 H4SiO4 (µg g ) 60.78 1.14 21.78 11.03 50.67

-1 NH3 (µg kg ) 5,077 18.85 324 587 181 N:P ratio (molar) 525,564 0.29 23,591 62,745 266 Total salt (µg g-1) 80,482 9.46 7,932 13,266 167 STD, standard deviation; CV, coefficient of variation

128

Table 4.2. Multiple linear regression and random forest model statistics

Out-of-the-bag (OOB) multiple linear regression and random forest model statistics generated in R.

Multiple Random forest regression Variance explained Most important Least important R2 p-value (%) variable variable F- 0.27 <<0.001 36 Elevation Distance from coast Cl- 0.05 0.082 20 Elevation Distance from coast Distance from NO - 0.18 <<0.001 -4 Distance from coast 3 glacier 2- SO4 0.37 <<0.001 44 Elevation Distance from coast

3- PO4 0.16 0.017 -7 Latitude Distance from coast

- ClO4 0.1 0.010 -3 Elevation Distance from coast

- ClO3 0.33 <<0.001 43 Latitude Distance from glacier Ca2+ 0.26 <<0.001 46 Soil moisture Distance from coast Mg2+ 0.29 <<0.001 22 Elevation Distance from coast Na+ 0.21 <<0.001 38 Elevation Distance from coast K+ 0.4 <<0.001 62 Elevation Distance from coast Sr2+ 0.55 <<0.001 62 Elevation Distance from coast

NH3 0.29 <<0.001 54 Elevation Distance from glacier Distance from N:P 0.37 <<0.001 -3 Distance from coast glacier Total 0.37 <<0.001 45 Elevation Distance from coast salts

129

Table 4.3. Multiple linear regression and random forest statistics between predicted and measured concentrations

Multiple linear regression and random forest statistics between predicted and measured concentrations from the Shackleton and Darwin Glacier regions. R2 and p-values are reported for the correlations between measured and predicted concentrations. Regression slopes and test statistic values (t) were calculated using the smatr library (Warton et al.,

2012) in R to evaluate the null hypothesis (H0) of slope = 1. Higher test statistic values (closer to one) indicate that we reject the null hypothesis.

Multiple Linear Regression Random Forest Analyte R2 p-value Reg. Test statistic (t) R2 p-value Reg. Test statistic (t) slope H0 slope = 1 slope H0 slope = 1 Shackleton Mg2+ 0.52 <<0.001 0.52 -0.711 0.79 <<0.001 0.58 0.780 Sr2+ 0.39 <0.001 1.22 0.247* 0.67 <<0.001 0.91 -0.166**

Total 0.46 <<0.001 0.76 -0.343* 0.51 <<0.001 0.93 -0.107** salts Ca2+ 0.25 0.004 0.42 -0.747 0.49 <<0.001 0.61 -0.586

2- SO4 0.43 <<0.001 1.07 0.093** 0.45 <<0.001 1.10 0.130** K+ 0.42 <<0.001 1.54 0.504* 0.42 <<0.001 1.79 0.629 F- 0.50 <<0.001 1.22 0.267* 0.42 <0.001 1.78 0.617 N:P 0.05 0.241 0.59 -0.517 0.42 <<0.001 0.35 -0.867 ratio Cl- 0.07 0.144 0.28 -0.867 0.41 <<0.001 0.70 -0.424*

- ClO4 0.35 <0.001 2.01 0.685 0.40 <0.001 3.40 0.897

NH3 0.11 0.070 1.04 0.037** 0.36 <0.001 1.09 0.106** Na+ 0.33 <0.001 0.91 -0.112** 0.31 0.001 1.54 0.473* - NO3 0.40 <0.001 0.47 -0.725 0.29 0.002 0.56 -0.594

- ClO3 0.13 0.043 1.20 0.197** 0.28 0.002 0.71 -0.382*

3- PO4 0.18 0.016 0.50 -0.645 0.20 0.022 0.15 -0.967 Darwin Mg2+ - - - - 0.87 <<0.001 0.39 -0.948 K+ - - - - 0.87 <<0.001 0.49 -0.895 Cl- - - - - 0.57 0.011 0.13 -0.984

Continued

130

Table 4.3 continued Total - - - - 0.44 0.001 3.25 0.940 salts Na+ - - - - 0.34 0.078 0.23 -0.931

- NO3 - - - - 0.33 0.080 0.65 -0.476* Ca2+ - - - - 0.29 0.110 0.17 -0.961 N:P - - - - 0.01 0.765 8.04 0.970 ratio * t < 0.5; ** t < 0.20

131

Table 4.4. Multiple linear regression and random forest model (MAE) and (RMSE)

Multiple linear regression and random forest model mean absolute error (MAE) and root mean square error (RMSE) between measured and predicted values for the Shackleton and Darwin Glacier regions.

Multiple Linear Regression Random Forest Analyte MAE RMSE MAE RMSE Shackleton Mg2+ 300 461 204 347 Sr2+ 3.74 4.96 1.83 2.90 Total salts 5,638 7,067 4,396 7,027 Ca2+ 797 1,102 554 912 2- SO4 3,314 3,889 2,197 3,780 K+ 15.86 21.16 13.48 25.61 F- 3.14 4.19 3.13 6.31 N:P ratio 39,668 59,322 7,307 17,206 Cl- 936 1,535 658 1,241 - ClO4 1,183 1,563 875 2,963

NH3 214 301 158 244 Na+ 883 1,166 918 1,727 - NO3 1,198 1,908 1,133 2,035 - ClO3 1,107 1,631 343 1,054 3- PO4 428 690 261 742 Darwin Mg2+ 6,297 6,315 302 475 K+ 1,056 1,056 13.33 15.84 Cl- 205,495 205,529 2,139 3,325 Total salts 214,856 214,933 5,541 7,589 Na+ 8,334 8,533 1,499 2,597

- NO3 128,322 128,371 3,256 4,866 Ca2+ 70,276 70,297 1,406 2,074 N:P ratio 18,100,000 18,100,000 18,717 46,939 MAE, mean absolute error; RMSE, root mean squared error

132

Figure 4.1. Map of 220 soil sample locations in the Shackleton Glacier region

Samples were collected and analyzed from the exposed soils along the Shackleton Glacier, a major outlet glacier of the EAIS (a), in three zones. The upper zone (b) was located at the head of Shackleton Glacier, the middle zone (c) was the central portion, and the lower zone (d) was at the mouth of the glacier where it drains into the Ross Sea. Satellite images were provided courtesy of the Polar Geospatial Center (PGC).

133

- - Figure 4.2. Total water-soluble salts, water-soluble N:P molar ratio, and ClO4 and ClO3 compared to elevation, distance from the coast, and distance from the glacier

- - Total water-soluble salts, water-soluble N:P molar ratio, and ClO4 and ClO3 concentrations (log scale) were compared to elevation, distance from the coast, and distance from the glacier for samples from the three geographic zones. Linear regression lines are plotted and R2 values are reported for each relationship. The horizontal orange line represents nematode salt tolerance of ~2,600 (Nkem et al., 2006) and the green line represents the Redfield ratio, N:P = 16 for phytoplankton in the ocean. 134

Figure 4.3. Anion and cation ternary diagrams for water-soluble salts

Anion and cation ternary diagrams for the three geographic zones.

135

Figure 4.4. Principal component analysis (PCA) biplot with all anions, cations, nutrients, and soil moisture

Principal component analysis (PCA) biplot generated in R using factoextra and built in software libraries with all anions, cations, nutrients, and soil moisture for the three geographic zones. Principal component 1 and principal component 2 are plotted on the x and y axes, respectively. Shaded ellipses represent 95% confidence intervals.

136

Figure 4.5. Spearman’s rank correlation matrix

Spearman’s rank correlation matrix generated in R using the corrplot library. The colors represent correlation coefficients, indicating the strength and magnitude of the correlation. The blue box indicates the geographic variables and soil moisture, which were variables used in the multiple linear regression and random forest models.

137

138

Figure 4.6. Inverse distance weighted (IDW) interpolations of total salt concentration

Inverse distance weighted (IDW) interpolations of total salt concentration for Roberts Massif (a), Bennett Platform (b), and Thanksgiving Valley (c). The color scale represents the 10 natural breaks in the data. Interpolations were created and mapped using the Geostatistical Analyst tool in ArcMap 10.3.

Figure 4.7. R2 values for the multiple linear regression and random forest model predicted and measured values

R2 values for the multiple linear regression and random forest model predicted and measured values for the different analytes (Table 4.3). Test datasets include the Shackleton Glacier region (n =31) and the Darwin Glacier region (n = 10) (Magalhães et al., 2012). Analytes with slopes near 1, indicating good agreement between measured and predicted values are indicated (* t < 0.5; ** t < 0.20).

139

Chapter 5. Conclusions

The results of the three independent studies in this work reflect the complex relationship between geochemistry, geography, and glacial history. The measurements and data include some of the most southern latitude isotopic and geochemical data reported. Water-soluble salt geochemistry of soils from the Shackleton Glacier region is influenced by elevation, latitude, longitude, geology, surface exposure age, and availability of liquid water, and geochemical gradients reflecting these variables have been developed and discussed. In general, soils which have the lowest water-soluble N:P

- molar ratios (driven largely by atmospheric NO3 deposition), lowest total salt concentrations (driven by atmospheric deposition and chemical weathering), and lowest

- - ClO3 and ClO4 concentrations are those with the youngest relative exposure ages. While not directly addressing the biotic component of Antarctic soil environments, the findings and data summarized below have important implications and applications for determining locations for past refugia and understanding ecosystem habitat suitability in extreme terrestrial environments.

Summary of conclusions:

1. The most abundant water-soluble anion in high elevation Shackleton Glacier

2- - region soils is SO4 , contrary to previous studies which have found NO3 to be the

140

most abundant. This is likely due to the solubility of nitrate salts and other loss

terms not yet quantifiable (such as soil photolytic loss).

2. The chemical composition of water-soluble salts varies throughout the region and

is likely reflective of the current and past availability of water.

15 17 - 3. The δ N and Δ O isotopic composition of water-soluble NO3 in soil leaches

- indicates that NO3 is primarily derived from the atmosphere, with varying

contributions from the troposphere (0-70%) and stratosphere (30-100%).

34 18 2- 4. The δ S and δ O isotopic composition of water-soluble SO4 in soil leaches

2- suggests that SO4 is deposited as secondary atmospheric sulfate (SAS) and

derived from the oxidation of SO2, H2S, and/or dimethyl sulfide by H2O2,

carbonyl sulfide (COS), and ozone in the atmosphere. Pyrite weathering is a

secondary, minor source.

13 18 5. The δ C and δ O isotopic composition of bulk soil HCO3 + CO3 suggests

carbonate minerals were formed at the surface cryogenically from rapid

evaporation/sublimation or freezing of fluids.

6. Meteoric 10Be concentrations in Shackleton Glacier region soils included the

highest concentrations measured in Antarctica to date.

7. Calculated maximum exposure ages range from 4.1 Ma at Roberts Massif near the

Polar Plateau to 0.11 Ma at Bennett Platform further north. When corrected for

inheritance, the ages (representing a minimum) range from 0.14 Ma at Roberts

Massif to 0.04 Ma at Thanksgiving Valley.

141

8. Exposure ages are generally youngest at lower elevations, closer to the Ross Ice

Shelf, and closer to the Shackleton Glacier or other tributary glaciers.

- 10 9. NO3 and meteoric Be concentrations can be coupled to provide an efficient and

cost-effective method to estimate exposure ages.

10. The Shackleton Glacier was likely present in its current form since at least the

Pleistocene in southern portions of the region, while towards the glacier terminus,

the Shackleton Glacier is more susceptible to changes in climate and has likely

retreated in the past. Though speculative, other EAIS outlet glaciers probably

behaved similarly.

11. The geochemistry of Shackleton Glacier region soils can be separated into three

zones related to geography and glacial history: 1) an upper zone near the head of

the glacier which is characterized by high total water-soluble salt concentrations,

- - high water-soluble N:P ratios, and high ClO4 and ClO3 concentrations, 2) a lower

3- zone with low total salt concentrations and higher PO4 concentrations, and 3) a

middle zone with intermediate values.

12. Interpolations of total water-soluble salt concentrations reflect the local variability

of salt deposition and accumulation due to soil age and availability of liquid

water. The interpolations can be used to estimate salt concentration trends on a

smaller scale.

13. Latitude, longitude, elevation, distance from the coast, distance from the glacier,

all parameters currently (or soon to be) measurable remotely can be correlated

with surface soil geochemistry.

142

14. Machine learning algorithms, such as multiple linear regression and random

forest, can be used to predict/estimate geochemical gradients given the parameters

listed above. These results can be used to estimate which soils might be the most

suitable habitats for terrestrial organisms.

Recommendations for future work:

1) Attain meteoric 10Be concentration data for the remaining four depth profiles to

better constrain estimated ages. Additionally, analyze concentrations across entire

transects to estimate the LGM trim line.

2- 17 2) Analyze SO4 for Δ O to identify specific secondary atmospheric sulfate (SAS)

formation mechanisms.

- 3) Perform photochemical experiments on NO3 to determine rates of photolysis in

Antarctic soils.

4) Analyze more samples and attain more soil geochemical data to include in the

random forest model training dataset. A greater number of data points should

increase the model’s predictive capabilities.

5) Use the random forest model to create a greater density of geochemical data

points and create higher resolution interpolation maps for the Shackleton Glacier

region.

10 - 6) Investigate the relationship between meteoric Be and NO3 concentrations along

other outlet glaciers to determine if the relationship is well-established across the

CTAM.

143

Recommendations 2 and 3 would mainly entail additional analyses since the samples have already been collected. Recommendations 3-6 would require logistical support for sample collection and field experiments. All recommendations are in accordance with the roadmap for Antarctic and Southern Ocean science, which is an international collaboration for the future of Antarctic research (Kennicutt et al., 2015).

The horizon scan identifies pressing questions for Antarctic scientific exploration for the next two decades. The goals include understanding how ecosystems have responded, and may respond in the future, to climate transitions (Kennicutt et al., 2015), an issue particularly pertinent to the research discussed in this work.

144

References

Ackert, R. P., and Kurz, M. D. (2004). Age and uplift rates of Sirius Group sediments in the Dominion Range, Antarctica, from surface exposure dating and geomorphology. Glob. Planet. Change 42, 207–225. doi:10.1016/j.gloplacha.2004.02.001. Adams, B. J., Bardgett, R. D., Ayres, E., Wall, D. H., Aislabie, J., Bamforth, S., et al. (2006). Diversity and distribution of Victoria Land biota. Soil Biol. Biochem. 38, 3003–3018. doi:10.1016/j.soilbio.2006.04.030. Alexander, B., Thiemens, M. H., Farquhar, J., Kaufman, A. J., Savarino, J., and Delmas, R. J. (2003). East Antarctic ice core sulfur isotope measurements over a complete glacial-interglacial cycle. J. Geophys. Res. Atmos. 108, 4786. doi:10.1029/2003jd003513. Altabet, M. A., Wassenaar, L. I., Douence, C., and Roy, R. (2019). A Ti(III) reduction method for one‐step conversion of seawater and freshwater nitrate into N 2 O for stable isotopic analysis of 15 N/ 14 N, 18 O/ 16 O and 17 O/ 16 O. Rapid Commun. Mass Spectrom. 33, 1227–1239. doi:10.1002/rcm.8454. Anderson, J. B., Shipp, S. S., Lowe, A. L., Wellner, J. S., and Mosola, A. B. (2002). The Antarctic Ice Sheet during the Last Glacial Maximum and its subsequent retreat history: a review. Quat. Sci. Rev. 21, 49–70. doi:10.1016/S0277-3791(01)00083-X. Antipov, E. A., and Pokryshevskaya, E. B. (2012). Mass appraisal of residential apartments: An application of Random forest for valuation and a CART-based approach for model diagnostics. Expert Syst. Appl. 39, 1772–1778. doi:10.1016/j.eswa.2011.08.077. Augustin, L., Barbante, C., Barnes, P. R. F., Barnola, J. M., Bigler, M., Castellano, E., et al. (2004). Eight glacial cycles from an Antarctic ice core. Nature 429, 623–628. doi:10.1038/nature02599. Balci, N., Shanks, W. C., Mayer, B., and Mandernack, K. W. (2007). Oxygen and sulfur isotope systematics of sulfate produced by bacterial and abiotic oxidation of pyrite. Geochim. Cosmochim. Acta 71, 3796–3811. doi:10.1016/j.gca.2007.04.017. Balco, G. (2011). Contributions and unrealized potential contributions of cosmogenic- nuclide exposure dating to glacier chronology, 1990-2010. Quat. Sci. Rev. 30, 3–27. doi:10.1016/j.quascirev.2010.11.003. Balco, G. (2020). Technical note: A prototype transparent-middle-layer data management and analysis infrastructure for cosmogenic-nuclide exposure dating. Geochronol. Discuss. in review. doi:10.5194/gchron-2020-6. Balco, G., Stone, J. O. H., Sliwinski, M. G., and Todd, C. (2019). Features of the glacial history of the Transantarctic Mountains inferred from cosmogenic 26 Al, 10 Be and

145

21 Ne concentrations in bedrock surfaces. Antarct. Sci. 26, 708–723. doi:10.1017/S0954102014000261. Ball, B. A., Adams, B. J., Barrett, J. E., Wall, D. H., and Virginia, R. A. (2018). Soil biological responses to C, N and P fertilization in a polar desert of Antarctica. Soil Biol. Biochem. 122, 7–18. doi:10.1016/J.SOILBIO.2018.03.025. Balter, A., Bromley, G., Balco, G., Thomas, H., and Jackson, M. S. (2020). A 14.5 million-year record of East Antarctic Ice Sheet fluctuations from the central Transantarctic Mountains, 2 constrained with cosmogenic. Cryosph. Discuss. in review. doi:10.5194/tc-2020-57. Bao, H. (2015). Sulfate: A time capsule for Earth’s O. Chem. Geol. 395, 108–118. doi:10.1016/j.chemgeo.2014.11.025. Bao, H., Campbell, D. A., Bockheim, J. G., and Thiemens, M. H. (2000). Origins of sulphate in Antarctic dry-valley soils as deduced from anomalous 17O compositions. Nature 407, 499–502. doi:10.1038/35035054. Bao, H., and Marchant, D. R. (2006). Quantifying sulfate components and their variations in soils of the McMurdo Dry Valleys, Antarctica. J. Geophys. Res. 111, D16301. doi:10.1029/2005JD006669. Baroni, M., Savarino, J., Cole-dai, J., Rai, V. K., and Thiemens, M. H. (2008). Anomalous sulfur isotope compositions of volcanic sulfate over the last millennium in Antarctic ice cores. J. Geophys. Res. 113, 1–12. doi:10.1029/2008JD010185. Barrett, J. E., Virginia, R. A., Lyons, W. B., McKnight, D. M., Priscu, J. C., Doran, P. T., et al. (2007). Biogeochemical stoichiometry of Antarctic Dry Valley ecosystems. J. Geophys. Res. 112, G01010. doi:10.1029/2005JG000141. Barrett, J. E., Virginia, R. A., Parsons, A. N., and Wall, D. H. (2005). Potential Soil Organic Matter Turnover in Taylor Valley, Antarctica. Source Arctic, Antarct. Alp. Res. Publ. By Inst. Arct. Alp. Res. Arctic, Antarct. Alp. Res. 37010822. Barrett, J. E., Virginia, R. A., Wall, D. H., Cary, S. C., Adams, B. J., Hacker, A. L., et al. (2006). Co-variation in soil biodiversity and biogeochemistry in northern and southern Victoria Land, Antarctica. Antarct. Sci. 18, 535–548. doi:10.1017/S0954102006000587. Barrett, P. (2013). Resolving views on Antarctic Neogene glacial history-The Sirius debate Antarctic climate history View project. Earth Environ. Sci. Trans. R. Soc. Edinburgh 104, 31–53. doi:10.1017/S175569101300008X. Barrett, P. J., Adams, C. J., McIntosh, W. C., Swisher, C. C., and Wilson, G. S. (1992). Geochronological evidence supporting Antarctic deglaciation three million years ago. Nature 359, 816–818. Bennett, K. R., Hogg, I. D., Adams, B. J., and Hebert, P. D. N. (2016). High levels of intraspecific genetic divergences revealed for Antarctic springtails: evidence for small-scale isolation during Pleistocene glaciation. Biol. J. Linn. Soc. 119, 166–178. doi:10.1111/bij.12796. Bierman, P. R., Corbett, L. B., Graly, J. A., Neumann, T. A., Lini, A., Crosby, B. T., et al. (2014). Preservation of a Preglacial Landscape Under the Center of the Greenland Ice Sheet. Science (80-. ). 344, 402–405.

146

doi:10.4159/harvard.9780674430501.c21. Bishop, J. L., Englert, P. A. J., Patel, S., Tirsch, D., Roy, A. J., Koeberl, C., et al. (2015). Mineralogical analyses of surface sediments in the Antarctic Dry Valleys: coordinated analyses of Raman spectra, reflectance spectra and elemental abundances. Philos. Transations R. Soc. 372, 20140198. doi:10.1098/rsta.2014.0198. Bisson, K. M., Welch, K. A., Welch, S. A., Sheets, J. M., Lyons, W. B., and Levy & A, J. S. G. (2015). Patterns and Processes of Salt Efflorescences in the McMurdo region. Arctic, Antarct. Alp. Res. 47, 407–425. doi:10.1657/AAAR0014-024. Blunier, T., and Brook, E. J. (2001). Timing of millennial-scale climate change in antarctica and greenland during the last glacial period. Science (80-. ). 291, 109– 112. doi:10.1126/science.291.5501.109. Bockheim, J. G. (1990). Soil development rates in the Transantarctic Mountains. Geoderma 47, 59–77. doi:10.1016/0016-7061(90)90047-D. Bockheim, J. G. (2002). Landform and Soil Development in the McMurdo Dry Valleys, Antarctica: A Regional Synthesis. Arctic, Antarct. Alp. Res. 34, 308–317. doi:10.1080/15230430.2002.12003499. Bockheim, J. G. (2008). Functional diversity of soils along environmental gradients in the Ross Sea region, Antarctica. Geoderma 144, 32–42. doi:10.1016/j.geoderma.2007.10.014. Bockheim, J. G., and McLeod, M. (2013). Glacial geomorphology of the Victoria Valley System, Ross Sea Region, Antarctica. Geomorphology 193, 14–24. doi:10.1016/j.geomorph.2013.03.020. Bottos, E. M., Laughlin, D. C., Herbold, C. W., Lee, C. K., McDonald, I. R., and Cary, S. C. (2020). Abiotic factors influence patterns of bacterial diversity and community composition in the Dry Valleys of Antarctica. FEMS Microbiol. Ecol. 96, 42. doi:10.1093/femsec/fiaa042. Breiman, L. (2001). Random forests. Mach. Learn. 45, 5–32. doi:10.1023/A:1010933404324. Bromley, G. R. M., Hall, B. L., Stone, J. O., Conway, H., and Todd, C. E. (2010). Late Cenozoic deposits at Reedy Glacier, Transantarctic Mountains: implications for former thickness of the West Antarctic Ice Sheet. Quat. Sci. Rev. 29, 384–398. doi:10.1016/j.quascirev.2009.07.001. Brook, E. J., Brown, E. T., Kurz, M. D., Ackert, R. P., Raisbeck, G. M., and Yiou, F. (1995). Constraints on age, erosion, and uplift of Neogene glacial deposits in the Transantarctic Mountains determined from in situ cosmogenic 10Be and 26Al. Geology 23, 1063–1066. doi:10.1130/0091-7613(1995)023<1063:coaeau>2.3.co;2. Brook, E. J., Kurz, M. D., Ackert, R. P., Denton, G. H., Brown, E. T., Raisbeck, G. M., et al. (1993). Chronology of taylor glacier advances in Arena Valley, Antarctica, using in situ cosmogenic 3He and 10Be. Quat. Res. 39, 11–23. doi:10.1006/qres.1993.1002. Brown, E. T., Edmond, J. M., Raisbeck, G. M., Bourlès, D. L., Yiou, F., and Measures, C. I. (1992). Beryllium isotope geochemistry in tropical river basins. Geochim.

147

Cosmochim. Acta 56, 1607–1624. doi:10.1016/0016-7037(92)90228-B. Brühl, C., Lelieveld, J., Crutzen, P. J., and Tost, H. (2012). The role of carbonyl sulphide as a source of stratospheric sulphate aerosol and its impact on climate. Atmos. Chem. Phys. 12, 1239–1253. doi:10.5194/acp-12-1239-2012. Bruno, L. A., Baur, H., Graf, T., Schlüchter, C., Signer, P., and Wieler, R. (1997). Dating of Sirius Group tillites in the Antarctic Dry Valleys with cosmogenic 3He and 21Ne. doi:10.1016/s0012-821x(97)00003-4. Burgener, L. K., Huntington, K. W., Sletten, R., Watkins, J. M., Quade, J., and Hallet, B. (2018). Clumped isotope constraints on equilibrium carbonate formation and kinetic isotope effects in freezing soils. Geochim. Cosmochim. Acta 235, 402–430. doi:10.1016/j.gca.2018.06.006. Calhoun, J. A., and Chadson, R. J. (1991). Sulfur isotope measurements of submicrometer sulfate aerosol particles over the Pacific Ocean. Geophys. Res. Lett. 18, 1877–1880. Campbell, I. B., Claridge, G. G. C., Campbell, D. I., and Balks, M. R. (2013). “The Soil Environment of the Mcmurdo Dry Valleys, Antarctica,” in Ecosystem dynamics in a polar desert; the McMurdo Dry Valleys, Antarctica, ed. J. C. Priscu, 297–322. doi:10.1029/ar072p0297. Caruso, T., Hogg, I. D., Nielsen, U. N., Bottos, E. M., Lee, C. K., Hopkins, D. W., et al. (2019). Nematodes in a polar desert reveal the relative role of biotic interactions in the coexistence of soil animals. Commun. Biol. 2. doi:10.1038/s42003-018-0260-y. Cary, S. C., McDonald, I. R., Barrett, J. E., and Cowan, D. A. (2010). On the rocks: The microbiology of Antarctic Dry Valley soils. Nat. Rev. Microbiol. 8, 129–138. doi:10.1038/nrmicro2281. Claridge, G. G. C., and Campbell, I. B. (1968). Soils of the Shackleton glacier region, Queen Maud Range, Antarctica. New Zeal. J. Sci. 11, 171–218. Claridge, G. G. C., and Campbell, I. B. (1977). Salts in Antarctic soils, their distribution and relationship to soil processes. Soil Sci. 123, 377–384. Clark, B. C., and Van Hart, D. C. (1981). The salts of Mars. Icarus 45, 370–378. doi:10.1016/0019-1035(81)90041-5. Clark, I. D., and Lauriol, B. (1992). Kinetic enrichment of stable isotopes in cryogenic calcites. Chem. Geol. 102, 217–228. doi:10.1016/0009-2541(92)90157-Z. Clark, P. U., Dyke, A. S., Shakun, J. D., Carlson, A. E., Clark, J., Wohlfarth, B., et al. (2009). The Last Glacial Maximum. Science (80-. ). 325, 710–714. doi:10.1126/science.1172873. Collins, G. E., Hogg, I. D., Convey, P., Barnes, A. D., and McDonald, I. R. (2019). Spatial and Temporal Scales Matter When Assessing the Species and Genetic Diversity of Springtails (Collembola) in Antarctica. Front. Ecol. Evol. 7, 76. doi:10.3389/fevo.2019.00076. Convey, P., Chown, S. L., Clarke, A., Barnes, D. K. A., Bokhorst, S., Cummings, V., et al. (2014). The spatial structure of Antarctic biodiversity. Ecol. Monogr. 84, 203– 244. doi:10.1890/12-2216.1. Convey, P., Gibson, J. A. E., Hillenbrand, C. D., Hodgson, D. A., Pugh, P. J. A., Smellie,

148

J. L., et al. (2008). Antarctic terrestrial life - Challenging the history of the frozen continent? Biol. Rev. 83, 103–117. doi:10.1111/j.1469-185X.2008.00034.x. Courtright, E. M., Wall, D. H., and Virginia, R. A. (2001). Determining habitat suitability for soil invertebrates in an extreme environment: The McMurdo Dry Valleys, Antarctica. Antarct. Sci. 13, 9–17. doi:10.1017/S0954102001000037. Courty, M. A., Marlin, C., Dever, L., Tremblay, P., and Vachier, P. (1994). The properties, genesis and environmental significance of calcitic pendents from the High Arctic (Spitsbergen). Geoderma 61, 71–102. doi:10.1016/0016- 7061(94)90012-4. Davidson, A. D., Hamilton, M. J., Boyer, A. G., Brown, J. H., and Ceballos, G. (2009). Multiple ecological pathways to extinction in mammals. doi:10.1073/pnas.0901956106. DeConto, R. M., and Pollard, D. (2016). Contribution of Antarctica to past and future sea-level rise. Nature 531, 591–597. doi:10.1038/nature17145. Denton, G. H., Sugden, D. E., Marchant, D. R., Hall, B. L., and Wilch, T. I. (1993). East Antarctic Ice Sheet Sensitivity to Pliocene Climatic Change from a Dry Valleys Perspective. Source Geogr. Ann. Ser. A 75, 155–204. Diaz, M. A., Adams, B. J., Welch, K. A., Welch, S. A., Opiyo, S. O., Khan, A. L., et al. (2018). Aeolian Material Transport and Its Role in Landscape Connectivity in the McMurdo Dry Valleys, Antarctica. J. Geophys. Res. Earth Surf. 123, 3323–3337. doi:10.1029/2017JF004589. Diaz, M. A., Corbett, L. B., Bierman, P. R., Adams, B. J., Wall, D. H., Hogg, I. D., et al. (2020a). Relative terrestrial exposure ages inferred from meteoric 10Be and NO3- concentrations in soils along the Shackleton Glacier, Antarctica. Earth Surf. Dyn., in review. doi:https://doi.org/10.5194/esurf-2020-50. Diaz, M. A., Li, J., Michalski, G., Darrah, T. H., Adams, B. J., Wall, D. H., et al. (2020b). Stable isotopes of nitrate, sulfate, and carbonate in soils from the Transantarctic Mountains, Antarctica: A record of atmospheric deposition and chemical weathering. Front. Earth Sci., in press. doi:10.3389/feart.2020.00341. Dickinson, W. W., Schiller, M., Ditchburn, B. G., Graham, I. J., and Zondervan, A. (2012). Meteoric Be-10 from Sirius Group suggests high elevation McMurdo Dry Valleys permanently frozen since 6 Ma. Earth Planet. Sci. Lett. 355–356, 13–19. doi:10.1016/j.epsl.2012.09.003. Dolan, A. M., De Boer, B., Bernales, J., Hill, D. J., and Haywood, A. M. (2018). High climate model dependency of Pliocene Antarctic ice-sheet predictions. Nat. Commun. 9, 1–12. doi:10.1038/s41467-018-05179-4. Doran, P. T., McKay, C. P., Clow, G. D., Dana, G. L., Fountain, A. G., Nylen, T., et al. (2002). Valley floor climate observations from the McMurdo dry valleys, Antarctica, 1986-2000. J. Geophys. Res. Atmos. 107, 1–12. doi:10.1029/2001JD002045. Dragone, N. B., Diaz, M. A., Hogg, I., Lyons, W. B., Jackson, W. A., Wall, D. H., et al. (2020). Exploring the boundaries of microbial habitability in soil. Proc. Natl. Acad. Sci., in review.

149

Dwyer, G. S., and Chandler, M. A. (2009). Mid-Pliocene sea level and continental ice volume based on coupled benthic Mg/Ca palaeotemperatures and oxygen isotopes. Philos. Trans. R. Soc. A Math. Phys. Eng. Sci. 367, 157–168. doi:10.1098/rsta.2008.0222. Ebert, K., Willenbring, J., Norton, K. P., Hall, A., and Hättestrand, C. (2012). Meteoric 10 Be concentrations from saprolite and till in northern Sweden: Implications for glacial erosion and age. doi:10.1016/j.quageo.2012.05.005. Elliot, D. H., and Fanning, C. M. (2008). Detrital zircons from upper Permian and lower Triassic Victoria Group sandstones, Shackleton Glacier region, Antarctica: Evidence for multiple sources along the Gondwana plate margin. Gondwana Res. 13, 259– 274. doi:10.1016/j.gr.2007.05.003. Erbland, J., Savarino, J., Morin, S., France, J. L., Frey, M. M., and King, M. D. (2015). Air-snow transfer of nitrate on the East Antarctic Plateau-Part 2: An isotopic model for the interpretation of deep ice-core records. Atmos. Chem. Phys 15, 12079–12113. doi:10.5194/acp-15-12079-2015. Faure, G., and Felder, R. P. (1981). Isotopic composition of stronium and sulfur in secondary gypsum crystals, Brown Hills, Transantarctic Mountains. J. Geochemical Explor. 14, 265–270. doi:10.1038/201599b0. Fountain, A. G., Lyons, W. B., Burkins, M. B., Dana, G. L., Doran, P. T., Lewis, K. J., et al. (2010). Physical Controls on the Taylor Valley Ecosystem, Antarctica. Bioscience 49, 961–971. Fraser, C. I., Nikula, R., Ruzzante, D. E., and Waters, J. M. (2012). Poleward bound: Biological impacts of Southern Hemisphere glaciation. Trends Ecol. Evol. 27, 462– 471. doi:10.1016/j.tree.2012.04.011. Freckman, D. W., and Virginia, R. A. (1998). “Soil Biodiversity and Community Structure in the Mcmurdo Dry Valleys, Antarctica,” in Ecosystem dynamics in a polar desert; the McMurdo dry valleys, Antarctica, ed. J. C. Priscu (American Geophysical Union (AGU)), 323–335. doi:10.1029/AR072p0323. Frey, M. M., Savarino, J., Morin, S., Erbland, J., and Martins, J. M. F. (2009a). Photolysis imprint in the nitrate stable isotope signal in snow and atmosphere of East Antarctica and implications for reactive nitrogen cycling. Atmos. Chem. Phys. Atmos. Chem. Phys. 9, 8681–8696. doi:10.5194/acp-9-8681-2009. Frey, M. M., Savarino, J., Morin, S., Erbland, J., and Martins, J. M. F. (2009b). Photolysis imprint in the nitrate stable isotope signal in snow and atmosphere of East Antarctica and implications for reactive nitrogen cycling. doi:10.5194/acp-9- 8681-2009. Gasson, E., DeConto, R. M., Pollard, D., and Levy, R. H. (2016). Dynamic Antarctic ice sheet during the early to mid-Miocene. Proc. Natl. Acad. Sci. 113, 3459–3464. doi:10.1073/pnas.1516130113. Golledge, N. R., Fogwill, C. J., Mackintosh, A. N., and Buckley, K. M. (2012). Dynamics of the last glacial maximum Antarctic ice-sheet and its response to ocean forcing. Proc. Natl. Acad. Sci. U. S. A. 109, 16052–16056. doi:10.1073/pnas.1. Golledge, N. R., and Levy, R. H. (2011). Geometry and dynamics of an East Antarctic

150

Ice Sheet outlet glacier, under past and present climates. J. Geophys. Res. 116, F03025. doi:10.1029/2011JF002028. Golledge, N. R., Levy, R. H., McKay, R. M., Fogwill, C. J., White, D. A., Graham, A. G. C., et al. (2013a). Glaciology and geological signature of the Last Glacial Maximum Antarctic ice sheet. Quat. Sci. Rev. 78, 225–247. doi:10.1016/j.quascirev.2013.08.011. Golledge, N. R., Levy, R. H., McKay, R. M., Fogwill, C. J., White, D. A., Graham, A. G. C., et al. (2013b). Glaciology and geological signature of the Last Glacial Maximum Antarctic ice sheet. Quat. Sci. Rev. 78, 225–247. doi:10.1016/j.quascirev.2013.08.011. Gooseff, M. N., Berry Lyons, W., McKnight, D. M., Vaughn, B. H., Fountain, A. G., and Dowling, C. (2006). A Stable Isotopic Investigation of a Polar Desert Hydrologic System, McMurdo Dry Valleys, Antarctica. Arctic, Antarct. Alp. Res. 38, 60–71. doi:10.1657/1523-0430(2006)038[0060:ASIIOA]2.0.CO;2. Graham, I., Ditchburn, R. G., Claridge, G. G. G., Whitehead, N. E., Zondervan, A., and Sheppard, D. S. (2002). Dating Antarctic soils using atmospheric derived 10Be and nitrate. R. Soc. New Zeal. Bull. 35, 429–436. Graly, J. A., Bierman, P. R., Reusser, L. J., and Pavich, M. J. (2010). Meteoric 10Be in soil profiles - A global meta-analysis. Geochim. Cosmochim. Acta 74, 6814–6829. doi:10.1016/j.gca.2010.08.036. Graly, J. A., Licht, K. J., Druschel, G. K., and Kaplan, M. R. (2018). Polar desert chronologies through quantitative measurements of salt accumulation. Geology 46, 351–354. doi:10.1130/G39650.1. Gulick, S. P. S., Shevenell, A. E., Montelli, A., Fernandez, R., Smith, C., Warny, S., et al. (2017). Initiation and long-term instability of the East Antarctic Ice Sheet. Nature 552, 225–229. doi:10.1038/nature25026. Hagen, E. H., Koeberl, C., and Faure, G. (1990). “Extraterrestrial spherules in glacial sediment, Beardmore Glacier area, Transantarctic Mountains,” in Contributions to Antarctic Research I, Antarctic Research Series, ed. D. H. Elliot (Washington, DC: American Geophysical Union), 19–24. Hambrey, M. J., Webb, P. N., Harwood, D. M., and Krissek, L. A. (2003). Neogene glacial record from the Sirius Group of the Shackleton Glacier region, central Transantarctic Mountains, Antarctica. GSA Bull. 115, 994–1015. doi:10.1130/B25183.1. Heindel, R. C., Spickard, A. M., and Virginia, R. A. (2017). Landscape-scale soil phosphorus variability in the McMurdo Dry Valleys. Antarct. Sci. 29, 252–263. doi:10.1017/S0954102016000742. Heung, B., Bulmer, C. E., and Schmidt, M. G. (2014). Predictive soil parent material mapping at a regional-scale: A Random Forest approach. Geoderma 214–215, 141– 154. doi:10.1016/j.geoderma.2013.09.016. Higgins, S. M., Hendy, C. H., and Denton, G. H. (2000). Geochronology of Bonney Drift, Taylor Valley, Antarctica: Evidence for interglacial expansions of Taylor Glacier. Geogr. Ann. Ser. A Phys. Geogr. 82, 391–409. doi:10.1111/j.0435-

151

3676.2000.00130.x. Holser, W. T., and Kaplan, I. R. (1966). Isotope geochemistry of sedimentary sulfates. Chem. Geol. 1, 93–135. doi:https://doi.org/10.1016/0009-2541(66)90011-8. Honrath, R. E., Guo, S., Peterson, M. C., Dziobak, M. P., Dibb, J. E., and Arsenault, M. A. (2000). Photochemical production of gas phase NOx from ice crystal NO3-. J. Geophys. Res. Atmos. 105, 24183–24190. doi:10.1029/2000JD900361. Honrath, R. E., Peterson, M. C., Guo, S., Dibb, J. E., Shepson, P. B., and Campbell, B. (1999). Evidence of NOx Production Within or Upon Ice Particles in the Greenland Snowpack. Geophys. Res. Lett. 26, 695–698. Huybrechts, P. (1993). Glaciological Modelling of the Late Cenozoic East Antarctic Ice Sheet: Stability or Dynamism? Geogr. Ann. Ser. A, Phys. Geogr. 75, 221–238. doi:10.1080/04353676.1993.11880395. Ivy-Ochs, S., Schluchter, C., Kubik, P. W., Dittrich-Hannen, B., and Beer, J. (1995). Minimum 10Be exposure ages of early Pliocene for the Table Mountain plateau and the Sirius Group at Mount Fleming, Dry Valleys, Antarctica. Geology 23, 1007– 1010. Jackson, A., Davila, A. F., Böhlke, J. K., Sturchio, N. C., Sevanthi, R., Estrada, N., et al. (2016). Deposition, accumulation, and alteration of Cl-, NO3-, ClO4- and ClO3- salts in a hyper-arid polar environment: Mass balance and isotopic constraints. Geochim. Cosmochim. Acta 182, 197–215. doi:10.1016/j.gca.2016.03.012. Jackson, M. S., Hall, B. L., and Denton, G. H. (2018). Asynchronous behavior of the Antarctic Ice Sheet and local glaciers during and since Termination 1, Salmon Valley, Antarctica. Earth Planet. Sci. Lett. 482, 396–406. doi:10.1016/j.epsl.2017.11.038. Jackson, W. A., Böhlke, J. K., Andraski, B. J., Fahlquist, L., Bexfield, L., Eckardt, F. D., et al. (2015). Global patterns and environmental controls of perchlorate and nitrate co-occurrence in arid and semi-arid environments. Geochim. Cosmochim. Acta 164, 502–522. doi:10.1016/J.GCA.2015.05.016. Jackson, W. A., Davila, A. F., Estrada, N., Lyons, W. B., Coates, J. D., and Priscu, J. C. (2012). Perchlorate and chlorate biogeochemistry in ice-covered lakes of the McMurdo Dry Valleys, Antarctica. Geochim. Cosmochim. Acta 98, 19–30. doi:10.1016/j.gca.2012.09.014. Jones, A. E., Weller, R., Anderson, P. S., Jacobi, H. W., Wolff, E. W., Schrems, O., et al. (2001). Measurements of NOx emissions from the Antarctic snowpack. Geophys. Res. Lett. 28, 1499–1502. doi:10.1029/2000GL011956. Jones, L. M., and Faure, G. (1967). Origin of the salts in Lake Vanda, Wright Valley, Southern Victoria Land, Antarctica. Earth Planet. Sci. Lett. 3, 101–106. doi:10.1016/0012-821x(67)90019-2. Jones, R. S., Mackintosh, A. N., Norton, K. P., Golledge, N. R., Fogwill, C. J., Kubik, P. W., et al. (2015). Rapid Holocene thinning of an East Antarctic outlet glacier driven by marine ice sheet instability. Nat. Commun. 6, 9910. doi:10.1038/ncomms9910. Jonsell, U., Hansson, M. E., Mörth, C.-M., and Torssander, P. (2005). Sulfur isotopic signals in two shallow ice cores from Dronning Maud Land, Antarctica. Tellus B

152

Chem. Phys. Meteorol. 57, 341–350. doi:10.3402/tellusb.v57i4.16558. Jouzel, J., and Souchez, R. A. (1982). Melting- refreezing at the glacier sole and the isotopic composition of the ice. J. Glaciol. 28, 35–42. doi:10.1017/S0022143000011771. Kaplan, M. R., Licht, K. J., Winckler, G., Schaefer, J. M., Bader, N., Mathieson, C., et al. (2017). Middle to Late Pleistocene stability of the central East Antarctic Ice Sheet at the head of Law Glacier. Geology 45, 963–966. doi:10.1130/G39189.1. Kennicutt, M. C., Chown, S. L., Cassano, J. J., Liggett, D., Peck, L. S., Massom, R., et al. (2015). A roadmap for Antarctic and Southern Ocean science for the next two decades and beyond. Antarct. Sci. 27, 3–18. doi:10.1017/S0954102014000674. Keys, J. R., and Williams, K. (1981). Origin of crystalline, cold desert salts in the McMurdo region, Antarctica. Geochim. Cosmochim. Acta 45, 2299–2309. doi:10.1016/0016-7037(81)90084-3. Kirkwood, C., Cave, M., Beamish, D., Grebby, S., and Ferreira, A. (2016). A machine learning approach to geochemical mapping. J. Geochemical Explor. 167, 49–61. doi:10.1016/j.gexplo.2016.05.003. Kunasek, S. A., Alexander, B., Steig, E. J., Sofen, E. D., Jackson, T. L., Thiemens, M. H., et al. (2010). Sulfate sources and oxidation chemistry over the past 230 years from sulfur and oxygen isotopes of sulfate in a West Antarctic ice core. J. Geophys. Res. 115, D18313. doi:10.1029/2010JD013846. Lacelle, D. (2007). Environmental setting, (micro) morphologies and stable C–O isotope composition of cold climate carbonate precipitates — a review and evaluation of their potential as paleoclimatic proxies. Quat. Sci. Rev. 26, 1670–1689. doi:10.1016/j.quascirev.2007.03.011. Lacelle, D., Lauriol, B., and Clark, I. D. (2006). Effect of chemical composition of water on the oxygen-18 and carbon-13 signature preserved in cryogenic carbonates, Arctic Canada: Implications in paleoclimatic studies. Chem. Geol. 234, 1–16. doi:10.1016/j.chemgeo.2006.04.001. Lacelle, D., Lauriol, B., and Clark, I. D. (2007). Origin, age, and palaeoenvironmental significance of carbonate precipitates from a granitic environment, Akshayuk Pass, southern Baffin Island, Canada. Can. J. Earth Sci. 44, 61–79. doi:10.1139/E06-088. LaPrade, K. E. (1984). Climate, geomorphology, and glaciology of the Shackleton Glacier area, Queen Maud Mountains, Transantarctic Mountains, Antarctica. Antarct. Res. Ser. 36, 163–196. doi:10.1029/ar036p0163. Legrand, M., Feniet-Saigne, C., and Saltzmant, E. S. (1991). Ice-core record of oceanic emissions of dimethylsulphide during the last climate cycle. Legrand, M. R., and Delmas, R. J. (1984). The ionic balance of Antarctic snow: A 10- year detailed record. Atmos. Environ. 18, 1867–1874. doi:10.1016/0004- 6981(84)90363-9. Levy, J. (2013). How big are the McMurdo Dry Valleys? Estimating ice-free area using Landsat image data. Antarct. Sci. 25, 119–120. doi:10.1017/S0954102012000727. Levy, J. S., Fountain, A. G., Welch, K. A., and Lyons, W. B. (2012). Hypersaline “wet patches” in Taylor Valley, Antarctica. Geophys. Res. Lett. 39, n/a-n/a.

153

doi:10.1029/2012GL050898. Lewis, A. R., Marchant, D. R., Ashworth, A. C., Hedenäs, L., Hemming, S. R., Johnson, J. V., et al. (2008). Mid-Miocene cooling and the extinction of tundra in continental Antarctica. Proc. Natl. Acad. Sci. U. S. A. 105, 10676–10680. doi:10.1073/pnas.0802501105. Lyons, W. B., Deuerling, K., Welch, K. A., Welch, S. A., Michalski, G., Walters, W. W., et al. (2016). The Soil Geochemistry in the Beardmore Glacier Region, Antarctica: Implications for Terrestrial Ecosystem History. Sci. Rep. 6, 26189. doi:10.1038/srep26189. Lyons, W. B., Foley, K. K., Carey, A. E., Diaz, M. A., Bowen, G. J., and Cerling, T. (2020). The isotopic geochemistry of CaCO3 encrustations in Taylor Valley, Antarctica: Implications for their origin. Acta Geogr. Slov. 60, 105–119. Lyons, W. B., Welch, K. A., Neumann, K., Toxey, J. K., McArthur, R., Williams, C., et al. (1998). Geochemical linkages among glaciers, streams and lakes within the Taylor Valley, Antarctica. Ecosyst. Dyn. a polar desert; McMurdo dry Val. Antarct. 72, 77–92. doi:10.1029/AR072p0077. MacKintosh, A., Golledge, N., Domack, E., Dunbar, R., Leventer, A., White, D., et al. (2011). Retreat of the East Antarctic ice sheet during the last glacial termination. Nat. Geosci. 4, 195–202. doi:10.1038/ngeo1061. Mackintosh, A. N., Verleyen, E., O’Brien, P. E., White, D. A., Jones, R. S., McKay, R., et al. (2014). Retreat history of the East Antarctic Ice Sheet since the Last Glacial Maximum. Quat. Sci. Rev. 100, 10–30. doi:10.1016/j.quascirev.2013.07.024. Magalhães, C., Stevens, M. I., Cary, S. C., Ball, B. A., Storey, B. C., Wall, D. H., et al. (2012). At Limits of Life: Multidisciplinary Insights Reveal Environmental Constraints on Biotic Diversity in Continental Antarctica. PLoS One 7, e44578. doi:10.1371/journal.pone.0044578. Marchant, D. R., and Denton, G. H. (1996). Miocene and Pliocene paleoclimate of the Dry Valleys region, Southern Victoria land: a geomorphological approach. Mar. Micropaleontol. Micropaleontol. 27, 253–27. doi:https://doi.org/10.1016/0377- 8398(95)00065-8. Marchant, D. R., Denton, G. H., Bockheim, J. G., Wilson, S. C., and Kerr, A. R. (1994). Quaternary changes in level of the upper Taylor Glacier, Antarctica: implications for paleoclimate and East Antarctic Ice Sheet dynamics. Boreas 23, 29–43. doi:10.1111/j.1502-3885.1994.tb00583.x. Marchant, D. R., Denton, G. H., Swisher, C. C., and Potter, N. (1996). Late Cenozoic Antarctic paleoclimate reconstructed from volcanic ashes in the Dry Valleys region of southern Victoria Land. Geol. Soc. Am. Bull. 108, 181–194. doi:https://doi.org/10.1130/0016-7606(1996)108%3C0181:LCAPRF%3E2.3.CO;2. Margerison, H. R., Phillips, W. M., Stuart, F. M., and Sugden, D. E. (2005). Cosmogenic 3He concentrations in ancient flood deposits from the Coombs Hills, northern Dry Valleys, East Antarctica: interpreting exposure ages and erosion rates. Earth Planet. Sci. Lett. 230, 163–175. doi:10.1016/J.EPSL.2004.11.007. Mayewski, P. A., and Goldthwait, R. P. (1985). Glacial events in the Transantarctic

154

Mountains: A record of the East Antarctic ice sheet. Antarct. Res. Ser. 36, 275–324. doi:10.1029/ar036p0275. Mayewski, P. A., Twickler, M. S., Lyons, W. B., Spencer, M. J., Meese, D. A., Gow, A. J., et al. (1990). The Dominion Range ice core, Queen Maud Mountains, Antarctica - general site and core characteristics with implications. J. Glaciol. 36, 11–16. doi:10.1017/S0022143000005499. McHargue, L. R., and Damon, P. E. (1991). The global beryllium 10 cycle. Rev. Geophys. 29, 141–158. doi:10.1029/91RG00072. McKay, R., Browne, G., Carter, L., Cowan, E., Dunbar, G., Krissek, L., et al. (2009). The stratigraphic signature of the late Cenozoic Antarctic Ice Sheets in the Ross Embayment. Bull. Geol. Soc. Am. 121, 1537–1561. doi:10.1130/B26540.1. McKay, R. M., Dunbar, G. B., Naish, T. R., Barrett, P. J., Carter, L., and Harper, M. (2008). Retreat history of the Ross Ice Sheet (Shelf) since the Last Glacial Maximum from deep-basin sediment cores around Ross Island. Palaeogeogr. Palaeoclimatol. Palaeoecol. 260, 245–261. doi:10.1016/j.palaeo.2007.08.015. Michalski, G., Bockheim, J. G., Kendall, C., and Thiemens, M. (2005). Isotopic composition of Antarctic Dry Valley nitrate: Implications for NOy sources and cycling in Antarctica. Geophys. Res. Lett. 32, 1–4. doi:10.1029/2004GL022121. Michalski, G., Böhlke, J. K., and Thiemens, M. (2004). Long term atmospheric deposition as the source of nitrate and other salts in the Atacama Desert, Chile: New evidence from mass-independent oxygen isotopic compositions. Geochim. Cosmochim. Acta 68, 4023–4038. doi:10.1016/j.gca.2004.04.009. Michalski, G., Scott, Z., Kabiling, M., and Thiemens, M. H. (2003). First measurements and modeling of Δ17O in atmospheric nitrate. Geophys. Res. Lett. 30, 1870. doi:10.1029/2003GL017015. Moore, H. (1977). The isotopic composition of ammonia, nitrogen dioxide and nitrate in the atmosphere. Atmos. Environ. 11, 1239–1243. doi:10.1016/0004-6981(77)90102- 0. Moorhead, D. L., Doran, P. T., Fountain, A. G., Lyons, W. B., Mcknight, D. M., Priscu, J. C., et al. (1999). Ecological Legacies: Impacts on Ecosystems of the McMurdo Dry Valleys McMurdo Dry Valleys. Source Biosci. 49, 1009–1019. doi:10.1525/bisi.1999.49.12.1009. Morgan, D., Putkonen, J., Balco, G., and Stone, J. (2010). Quantifying regolith erosion rates with cosmogenic nuclides 10 Be and 26 Al in the McMurdo Dry Valleys, Antarctica. J. Geophys. Res. 115, F03037. doi:10.1029/2009JF001443. Morin, S., Savarino, J., Frey, M. M., Domine, F., Jacobi, H. W., Kaleschke, L., et al. (2009). Comprehensive isotopic composition of atmospheric nitrate in the Atlantic Ocean boundary layer from 65°S to 79°N. J. Geophys. Res. Atmos. 114, 1–19. doi:10.1029/2008JD010696. Muscari, G., de Zafra, R. L., and Smyshlyaev, S. (2003). Evolution of the NOy-N2O correlation in the Antarctic stratosphere during 1993 and 1995. J. Geophys. Res. D Atmos. 108. doi:10.1029/2002jd002871. Naish, T., Powell, R., Levy, R., Wilson, G., Scherer, R., Talarico, F., et al. (2009).

155

Obliquity-paced Pliocene West Antarctic ice sheet oscillations. Nature 458, 322– 328. doi:10.1038/nature07867. Nakada, M., and Lambeck, K. (1988). The melting history of the late Pleistocene Antarctic ice sheet. doi:10.1038/333036a0. Nakai, N., Wada, H., Kiyosu, Y., and Takimoto, M. (1975). Stable isotope of water and studies on the origin and geological history salts in the Lake Vanda area, Antarctica. Geochem. J. 9, 7–24. doi:https://doi.org/10.2343/geochemj.9.7. Neumann, K., Lyons, W. B., Priscu, J. C., Desmarais, D. J., and Welch, K. A. (2004). The carbon isotopic composition of dissolved inorganic carbon in perennially ice- covered Antarctic lakes: searching for a biogenic signature. Ann. Glaciol. 39, 1–7. doi:10.3189/172756404781814465. Nezat, C. A., Lyons, W. B., and Welch, K. A. (2001). Chemical weathering in streams of a polar desert (Taylor Valley, Antarctica). Geol. Soc. Am. Bull. 113, 1401–1408. doi:10.1130/0016-7606(2001)113<1401:CWISOA>2.0.CO;2. Nishiizumi, K., Imamura, M., Caffee, M. W., Southon, J. R., Finkel, R. C., and McAninch, J. (2007). Absolute calibration of 10Be AMS standards. Nucl. Instruments Methods Phys. Res. B 258, 403–413. doi:10.1016/j.nimb.2007.01.297. Nkem, J. N., Virginia, A. R. A., Barrett, A. J. E., Wall, D. H., and Li, A. G. (2006). Salt tolerance and survival thresholds for two species of Antarctic soil nematodes. Polar Biol. 29, 643–651. doi:10.1007/s00300-005-0101-6. Patel, J., Shah, S., Thakkar, P., and Kotecha, K. (2015). Predicting stock market index using fusion of machine learning techniques. Expert Syst. Appl. 42, 2162–2172. doi:10.1016/j.eswa.2014.10.031. Patris, N., Delmas, R. J., and Jouzel, J. (2000). Isotopic signatures of sulfur in shallow Antarctic ice cores. J. Geophys. Res. 105, 7071–7078. doi:10.1029/1999JD900974. Paulsen, T. S., Encarnación, J., and Grunow, A. M. (2004). Structure and timing of transpressional deformation in the Shackleton Glacier area, Ross orogen, Antarctica. J. Geol. Soc. London. 161, 1027–1038. doi:10.1144/0016-764903-040. Pavich, M. J., Brown, L., Harden, J., Klein, J., and Middleton, R. (1986). 10Be distribution in soils from Merced River terraces, California. Geochim. Cosmochim. Acta 50, 1727–1735. doi:10.1016/0016-7037(86)90134-1. Pavich, M. J., Brown, L., Klein, J., and Middleton, R. (1984). 10Be accumulation in a soil chronosequence. Earth Planet. Sci. Lett. 68, 198–204. doi:10.1016/0012- 821X(84)90151-1. Perez-Lopez, R., Bejar, M., Giner-Robles, J. L., Gonzalez-Casado, J. M., Martinez-Diaz, J. J., Paredes, C., et al. (2007). Active tectonics on Deception Island (West- Antarctica); a new approach by using the fractal anisotropy of lineaments, fault slip measurements and the caldera collapse shape. Open-File Rep. - U. S. Geol. Surv., Short Research Paper 086. doi:10.3133/of2007-1047.srp086. Pertierra, L. R., Hughes, K. A., Vega, G. C., and Olalla-Tárraga, M. A. (2017). High resolution spatial mapping of human footprint across antarctica and its implications for the strategic conservation of avifauna. PLoS One 12, 1–21. doi:10.1371/journal.pone.0168280.

156

Peters, J., De Baets, B., Verhoest, N. E. C., Samson, R., Degroeve, S., Becker, P. De, et al. (2007). Random forests as a tool for ecohydrological distribution modelling. Ecol. Modell. 207, 304–318. doi:10.1016/j.ecolmodel.2007.05.011. Pisapia, C., Chaussidon, M., Mustin, C., and Humbert, B. (2007). O and S isotopic composition of dissolved and attached oxidation products of pyrite by Acidithiobacillus ferrooxidans: Comparison with abiotic oxidations. Geochim. Cosmochim. Acta 71, 2474–2490. doi:10.1016/j.gca.2007.02.021. Pollard, D., and DeConto, R. M. (2009). Modelling West Antarctic ice sheet growth and collapse through the past five million years. Nature 458, 329–332. doi:10.1038/nature07809. Prasad, A. M., Iverson, L. R., and Liaw, A. (2006). Newer classification and regression tree techniques: Bagging and random forests for ecological prediction. Ecosystems 9, 181–199. doi:10.1007/s10021-005-0054-1. Prentice, M. L., Bockheim, J. G., Wilson, S. C., Burckle, L. H., Hodell, D. A., Schlüchter, C., et al. (1993). “Late Neogene Antarctic Glacial History: Evidence from Central Wright Valley,” in The Antarctic Paleoenvironment: A Perspective on Global Change: Part Two, Volume 60 (American Geophysical Union (AGU)), 207– 250. doi:10.1002/9781118668061.CH11. Pribil, M. J., Ridley, W. I., and Emsbo, P. (2015). Sulfate and sulfide sulfur isotopes (δ34S and δ33S) measured by solution and laser ablation MC-ICP-MS: An enhanced approach using external correction. Chem. Geol. 412, 99–106. doi:10.1016/j.chemgeo.2015.07.014. Pruett, L. E., Kreutz, K. J., Wadleigh, M., Mayewski, P. A., and Kurbatov, A. (2004). Sulfur Isotopic Measurements from a West Antarctic Ice Core: Implications for Sulfate Source and Transport. Ann. Glaciol. 39, 161–168. doi:https://doi.org/10.3189/172756404781814339. R Core Team (2020). R: A language and environment for statistical computing. Rech, J. A., Quade, J., and Hart, W. S. (2003). Isotopic evidence for the source of Ca and S in soil gypsum, anhydrite and calcite in the Atacama Desert, Chile. Geochim. Cosmochim. Acta 67, 575–586. doi:10.1016/S0016-7037(02)01175-4. Reich, M., and Bao, H. (2018). Nitrate deposits of the Atacama Desert: A marker of long- term hyperaridity. Elements 14, 251–256. doi:10.2138/gselements.14.4.251. Richter, I., Herbold, C. W., Lee, C. K., McDonald, I. R., Barrett, J. E., and Cary, S. C. (2014). Influence of soil properties on archaeal diversity and distribution in the McMurdo Dry Valleys, Antarctica. FEMS Microbiol. Ecol. 89, 347–359. doi:10.1111/1574-6941.12322. Savarino, J., Kaiser, J., Morin, S., Sigman, D. M., and Thiemens, M. H. (2007). Nitrogen and oxygen isotopic constraints on the origin of atmospheric nitrate in coastal Antarctica. Atmos. Chem. Phys 7, 1925–1945. doi:https://doi.org/10.5194/acp-7- 1925-2007. Savarino, J., Lee, C. C. W., and Thiemens, M. H. (2000). Laboratory oxygen isotopic study of sulfur (IV) oxidation: Origin of the mass-independent oxygen isotopic anomaly in atmospheric sulfates and sulfate mineral deposits on Earth. J. Geophys.

157

Res. 105, 29079–29088. Scherer, R. P., Aldahan, A., Tulaczyk, S., Possnert, G., Engelhardt, H., and Kamb, B. (1998). Pleistocene collapse of the West Antarctic ice sheet. Science (80-. ). 281, 82–85. doi:10.1126/science.281.5373.82. Scherer, R. P., DeConto, R. M., Pollard, D., and Alley, R. B. (2016). Windblown Pliocene diatoms and East Antarctic Ice Sheet retreat. Nat. Commun. 7, 1–9. doi:10.1038/ncomms12957. Schiller, M., Dickinson, W., Ditchburn, R. G., Graham, I. J., and Zondervan, A. (2009). Atmospheric 10 Be in an Antarctic soil: Implications for climate change. J. Geophys. Res. 114, 1–8. doi:10.1029/2008jf001052. Shaheen, R., Abauanza, M., Jackson, T. L., Mccabe, J., Savarino, J., and Thiemens, M. H. (2013). Tales of volcanoes and El-Niño southern oscillations with the oxygen isotope anomaly of sulfate aerosol. Proc. Natl. Acad. Sci. 110, 17662–17667. doi:10.1073/pnas.1213149110. Shakun, J. D., Corbett, L. B., Bierman, P. R., Underwood, K., Rizzo, D. M., Zimmerman, S. R., et al. (2018). Minimal East Antarctic Ice Sheet retreat onto land during the past eight million years. Nature 558, 284–287. doi:10.1038/s41586-018-0155-6. Skidmore, M., Sharp, M., and Tranter, M. (2004). Kinetic isotopic fractionation during carbonate dissolution in laboratory experiments: Implications for detection of microbial CO2 signatures using δ13C-DIC. Geochim. Cosmochim. Acta 68, 4309– 4317. doi:10.1016/j.gca.2003.09.024. Spector, P., Stone, J., Cowdery, S. G., Hall, B., Conway, H., and Bromley, G. (2017). Rapid early-Holocene deglaciation in the Ross Sea, Antarctica. Geophys. Res. Lett. 44, 7817–7825. doi:10.1002/2017GL074216. Stafoggia, M., Bellander, T., Bucci, S., Davoli, M., de Hoogh, K., de’ Donato, F., et al. (2019). Estimation of daily PM10 and PM2.5 concentrations in Italy, 2013–2015, using a spatiotemporal land-use random-forest model. Environ. Int. 124, 170–179. doi:10.1016/j.envint.2019.01.016. Steig, E., Stuiver, M., and Polissar, P. (1995). Cosmogenic isotope concentrations at Taylor Dome, Antarctica. Antarct. J. United States 30, 95–97. Stevens, M. I., Greenslade, P., Hogg, I. D., and Sunnucks, P. (2006). Southern Hemisphere Springtails: Could Any Have Survived Glaciation of Antarctica? Mol. Biol. Evol. 23, 874–882. doi:10.1093/molbev/msj073. Stevens, M. I., and Hogg, I. D. (2003). Long-term isolation and recent range expansion from glacial refugia revealed for the endemic springtail Gomphiocephalus hodgsoni from Victoria Land, Antarctica. Mol. Ecol. 12, 2357–2369. doi:10.1046/j.1365- 294X.2003.01907.x. Stone, J. (1998). A rapid fusion method for separation of beryllium-10 from soils and silicates. Geochim. Cosmochim. Acta 62, 555–561. doi:10.1016/S0016- 7037(97)00340-2. Strasky, S., Di Nicola, L., Baroni, C., Salvatore, M. C., Baur, H., Kubik, P. W., et al. (2009). Surface exposure ages imply multiple low-amplitude Pleistocene variations in East Antarctic Ice Sheet, Ricker Hills, Victoria Land. Antarct. Sci. 21, 59–69.

158

doi:10.1017/S0954102008001478. Stroeven, A. P., Prentice, M. L., and Kleman, J. (1996). On marine microfossil transport and pathways in Antarctica during the late Neogene: Evidence from the Sirius Group at Mount Fleming. Geology 24, 727–730. doi:10.1130/0091- 7613(1996)024<0727:ommtap>2.3.co;2. Sugden, D. E. (1996). The East Antarctic Ice Sheet: Unstable Ice or Unstable Ideas? Trans. Inst. Br. Geogr. 21, 443. doi:10.2307/622590. Sugden, D. E., Marchant, D. R., and Denton, G. H. (1993). The case for a stable East Antarctic ice sheet. Geogr. Ann. Ser. A 75, 151–351. Sugden, D. E., Marchant, D. R., Potter, N., Souchez, R. A., Denton, G. H., Swisher, C. C., et al. (1995). Preservation of Miocene glacier ice in East Antarctica. Nature 376, 412–414. doi:10.1038/376412a0. Summerfield, M. A., Stuart, F. M., Cockburn, H. A. P., Sugden, D. E., Denton, G. H., Dunai, T., et al. (1999). Long-term rates of denudation in the Dry Valleys, Transantarctic Mountains, southern Victoria Land, Antarctica based on in-situ- produced cosmogenic 21Ne. Geomorphology 27, 113–129. doi:10.1016/S0169- 555X(98)00093-2. Sun, T., Socki, R. A., Bish, D. L., Harvey, R. P., Bao, H., Niles, P. B., et al. (2015). Lost cold Antarctic deserts inferred from unusual sulfate formation and isotope signatures. Nat. Commun. 6, 7579. doi:10.1038/ncomms8579. Szynkiewicz, A., Moore, C. H., Glamoclija, M., and Pratt, L. M. (2009). Sulfur isotope signatures in gypsiferous sediments of the Estancia and Tularosa Basins as indicators of sulfate sources, hydrological processes, and microbial activity. Geochim. Cosmochim. Acta 73, 6162–6186. doi:10.1016/j.gca.2009.07.009. Talarico, F. M., McKay, R. M., Powell, R. D., Sandroni, S., and Naish, T. (2012). Late Cenozoic oscillations of Antarctic ice sheets revealed by provenance of basement clasts and grain detrital modes in ANDRILL core AND-1B. Glob. Planet. Change 96–97, 23–40. doi:10.1016/j.gloplacha.2009.12.002. Tesoriero, A. J., Gronberg, J. A., Juckem, P. F., Miller, M. P., and Austin, B. P. (2017). Predicting redox-sensitive contaminant concentrations in groundwater using random forest classification. Water Resour. Res. 53, 7316–7331. doi:10.1002/2016WR020197. Toner, J. D., and Sletten, R. S. (2013). The formation of Ca-Cl-rich groundwaters in the Dry Valleys of Antarctica: Field measurements and modeling of reactive transport. Geochim. Cosmochim. Acta 110, 84–105. doi:10.1016/j.gca.2013.02.013. Toner, J. D., Sletten, R. S., and Prentice, M. L. (2013). Soluble salt accumulations in Taylor Valley, Antarctica: Implications for paleolakes and Ross Sea Ice Sheet dynamics. J. Geophys. Res. Earth Surf. 118, 198–215. doi:10.1029/2012JF002467. Tostevin, R., Turchyn, A. V., Farquhar, J., Johnston, D. T., Eldridge, D. L., Bishop, J. K. B., et al. (2014). Multiple sulfur isotope constraints on the modern sulfur cycle. Earth Planet. Sci. Lett. 396, 14–21. doi:10.1016/J.EPSL.2014.03.057. Uemura, R., Barkan, E., Abe, O., and Luz, B. (2010). Triple isotope composition of oxygen in atmospheric water vapor. Geophys. Res. Lett. 37, 1–5.

159

doi:10.1029/2009GL041960. Valletta, R. D., Willenbring, J. K., Lewis, A. R., Ashworth, A. C., and Caffee, M. (2015). Extreme decay of meteoric beryllium-10 as a proxy for persistent aridity. Sci. Rep. 5, 17813. doi:10.1038/srep17813. Vaniman, D. T., Bish, D. L., Chipera, S. J., Fialips, C. I., Carey, J. W., and Feldman, W. G. (2004). Magnesium sulphate salts and the history of water on Mars. Nature 431, 663–665. doi:10.1038/nature02973. Virginia, R. A., and Wall, D. H. (1999). How Soils Structure Communities in the Antarctic Dry Valleys. Bioscience 49, 973. doi:10.2307/1313731. Vogt, T., and Corte, A. E. (1996). Secondary precipitates in Pleistocene and present cryogenic environments (Mendoza Precordillera, Argentina, Transbaikalia, Siberia, and Seymour Island, Antarctica). Sedimentology 43, 53–64. doi:10.1111/j.1365- 3091.1996.tb01459.x. Wall, D., Bardgett, R. D., Behan-Pelletier, V., Herrick, J. ., Jones, H., Ritz, K., et al. eds. (2012). Soil Ecology and Ecosystem Services. Oxford: Oxford University Press. Walters, W. W., and Michalski, G. (2015). Theoretical calculation of nitrogen isotope equilibrium exchange fractionation factors for various NOy molecules. Geochim. Cosmochim. Acta 164, 284–297. doi:10.1016/j.gca.2015.05.029. Warton, D. I., Duursma, R. A., Falster, D. S., and Taskinen, S. (2012). smatr 3- an R package for estimation and inference about allometric lines. Methods Ecol. Evol. 3, 257–259. doi:10.1111/j.2041-210X.2011.00153.x. Webb, P. N., and Harwood, D. M. (1991). Late Cenozoic glacial history of the Ross embayment, Antarctica. Quat. Sci. Rev. 10, 215–223. doi:10.1016/0277- 3791(91)90020-U. Webb, P. N., Harwood, D. M., Mabin, M. G. C., and McKelvey, B. C. (1996). A marine and terrestrial Sirius Group succession, middle Beardmore Glacier-Queen Alexandra Range, Transantarctic Mountains, Antarctica. Mar. Micropaleontol. 27, 273–297. doi:10.1016/0377-8398(95)00066-6. Webb, P. N., Harwood, D. M., McKelvey, B. C., Mercer, J. H., and Stott, L. D. (1984). Cenozoic marine sedimentation and ice-volume variation on the East Antarctic craton. Geology 12, 287–291. doi:10.1130/0091- 7613(1984)12<287:cmsaiv>2.0.co;2. Webster-Brown, J., Gall, M., Gibson, J., Wood, S., and Hawes, I. (2010). The biogeochemistry of meltwater habitats in the Darwin Glacier region (80°S), Victoria Land, Antarctica. Antarct. Sci. 22, 646–661. doi:10.1017/S0954102010000787. Welch, K. A., Lyons, W. B., Whisner, C., Gardner, C. B., Gooseff, M. N., Mcknight, D. M., et al. (2010). Spatial variations in the geochemistry of glacial meltwater streams in the Taylor Valley, Antarctica. Antarct. Sci. 22, 662–672. doi:10.1017/S0954102010000702. Willenbring, J. K., and von Blanckenburg, F. (2010). Meteoric cosmogenic Beryllium-10 adsorbed to river sediment and soil: Applications for Earth-surface dynamics. Earth- Science Rev. 98, 105–122. doi:10.1016/j.earscirev.2009.10.008. Wilson, D. J., Bertram, R. A., Needham, E. F., van de Flierdt, T., Welsh, K. J., McKay,

160

R. M., et al. (2018). Ice loss from the East Antarctic Ice Sheet during late Pleistocene interglacials. Nature 561, 383–386. doi:10.1038/s41586-018-0501-8. Wilson, G. S. (1995). The neogene east antarctic ice sheet: A dynamic or stable feature? Quat. Sci. Rev. 14, 101–123. doi:10.1016/0277-3791(95)00002-7. Wynn-Williams, D. ., and Edwards, H. G. . (2000). Antarctic ecosystems as models for extraterrestrial surface habitats. Planet. Space Sci. 48, 1065–1075. doi:10.1016/S0032-0633(00)00080-5. You, C. F., Lee, T., and Li, Y. H. (1989). The partition of Be between soil and water. Chem. Geol. 77, 105–118. doi:10.1016/0009-2541(89)90136-8. Zeglin, L. H., Sinsabaugh, R. L., Barrett, J. E., Gooseff, M. N., and Takacs-Vesbach, C. D. (2009). Landscape Distribution of Microbial Activity in the McMurdo Dry Valleys: Linked Biotic Processes, Hydrology, and Geochemistry in a Cold Desert Ecosystem. Ecosystems 12, 562–573. doi:10.1007/s10021-009-9242-8.

161

Appendix A. Supplementary materials for Chapter 3

162

Figure A1. Max meteoric 10Be concentration versus 10Be inventory

Relationship between the measured maximum (or surface) meteoric 10Be concentration and the calculated inventory (Eq. 3.2). This relationship is used to infer 10Be inventories given a maximum or surface concentration (Graly et al., 2010).

163

Table A1. Soil grain size distribution of surface samples and depth profiles from Roberts Massif, Bennett Platform, and Thanksgiving Valley.

Sample Name % Gravel (>2 mm) % Sand (63-425 µm) % Silt (<63 µm) AV2-1 15.0 80.5 4.5 AV2-8 11.1 87.9 1.0 BP2-1, 0-5 34.0 61.4 4.6 BP2-1, 5-10 33.6 54.5 11.9 BP2-1, 10-15 37.8 46.2 16.0 BP2-8 15.6 83.2 1.1 MF2-1 32.1 65.8 2.1 MF2-4 34.4 64.8 0.8 MH2-1 36.0 62.0 2.1 MH2-8 31.6 67.3 1.1 MSP2-1 64.7 35.2 0.1 MSP2-4 33.6 66.1 0.3 MSP4-1 37.4 61.2 1.4 MW4-1 27.8 67.3 4.9 NP2-5 56.4 38.3 5.3 RM2-1, 0-5 13.2 69.3 17.5 RM2-1, 5-10 8.0 84.4 7.5 RM2-1, 10-15 7.5 80.8 11.7 RM2-8 24.9 68.5 6.7 SH3-2 15.7 77.3 7.0 SH3-8 4.7 92.2 3.2 TGV2-1, 0-5 27.7 71.6 0.7 TGV2-1, 5-10 32.4 66.7 0.9 TGV2-1, 10-15 44.1 54.7 1.2 TGV2-1, 15-20 29.3 69.3 1.4 TGV2-1, 20-25 21.6 76.8 1.6 TGV2-1, 25-30 52.2 45.1 2.7 TGV2-8 21.2 78.6 0.2 TN3-1 32.4 64.7 2.8 TN3-5 52.6 42.4 5.0

164

- 10 Table A2. NO3 concentrations and estimate of Be concentration from linear - 10 relationship between NO3 and Be.

- 5 -1 10 9 -1 Location Depth (cm) NO3 (10 µg kg ) Be estimate (10 atoms g ) Augustana 0 7.77 1.83 Augustana 5 12.2 1.97 Augustana 10 13.4 2.00 Schroeder 0 75.5 3.70 Schroeder 5 16.1 3.26 Schroeder 10 41.6 3.52 Franke 0 0.041 0.78 Franke 5 0.014 0.65 Franke 10 0.010 0.62 Franke 15 0.011 0.63 Roberts 0 6.94 4.57 Roberts 5 149 5.52 Roberts 10 30.7 5.01 Bennett 0 5.57 0.90 Bennett 5 39.8 0.34 Bennett 10 121 0.19 Thanksgiving 0 0.077 0.86 Thanksgiving 5 0.071 0.85 Thanksgiving 10 0.025 0.72 Thanksgiving 15 0.033 0.75 Thanksgiving 20 0.028 0.73 Thanksgiving 25 0.031 0.74 Heekin 0 18.0 2.10 Heekin 5 27.4 2.25 Heekin 10 18.8 2.11

165

Appendix B. Supplementary materials for Chapter 4

166

Table B1. Analytical precision and accuracy for water-soluble salts.

Analyte Precision Accuracy Standard (%) (%) F- 3.2 1.1 2015 USGS Interlaboratory Cl- 5.9 1.4 2015 USGS Interlaboratory - NO3 3.0 4.8 2015 USGS Interlaboratory 2- SO4 1.0 12.0 2015 USGS Interlaboratory 3- PO4 2.7 2.0 2015 USGS Interlaboratory - ClO4 19.3 10* Spike recovery - ClO3 15.4 10* Spike recovery Ca2+ 0.8 9.7 NIST1643e Mg2+ 0.8 1.6 NIST1643e Na+ 1.0 1.6 NIST1643e K+ 2.9 2.2 NIST1643e Sr2+ 2.8 0.9 NIST1643e

H4SiO4 0.8 0.9 2015 USGS Interlaboratory

NH3 19.2 0.2 2015 USGS Interlaboratory * Jackson et al. (2012)

167

Table B2. Water-soluble salt concentrations from 1:5 soil to water leaches from the Shackleton Glacier region.

Sample ID Location Zone Latitude Longitude Elevation (m) Distance from Distance from Soil moisture coast (km) glacier (m) (%) AV1-1 Augustana Upper -85.1762 -174.0885 1489 72 1108 0.82 AV1-2 Augustana Upper -85.1762 -174.0922 1492 72 1097 0.72 AV1-3 Augustana Upper -85.1763 -174.0943 1480 72 1086 1.02 AV1-4 Augustana Upper -85.1762 -174.0975 1473 72 1071 0.62 AV1-5 Augustana Upper -85.1761 -174.1005 1468 72 1049 0.90 AV1-6 Augustana Upper -85.1762 -174.1026 1455 72 1054 0.94 AV1-7 Augustana Upper -85.1762 -174.1071 1451 72 1031 0.59 AV1-8 Augustana Upper -85.1760 -174.1165 1433 72 993 0.80

168 AV2-1 Augustana Upper -85.1706 -174.1338 1410 72 388 1.04

AV2-2 Augustana Upper -85.1703 -174.1350 1405 72 352 1.60 AV2-3 Augustana Upper -85.1701 -174.1349 1390 72 329 0.98 AV2-4 Augustana Upper -85.1696 -174.1364 1369 72 275 1.11 AV2-5 Augustana Upper -85.1691 -174.1372 1388 72 226 1.46 AV2-6 Augustana Upper -85.1689 -174.1374 1376 72 195 1.14 AV2-7 Augustana Upper -85.1680 -174.1389 1374 72 104 0.64 AV2-8 Augustana Upper -85.1676 -174.1393 1378 72 61 0.73 AV3-1 Augustana Upper -85.1707 -174.1483 1439 72 392 0.63 AV3-2 Augustana Upper -85.1703 -174.1475 1430 72 352 0.54 AV3-3 Augustana Upper -85.1694 -174.1462 1416 72 251 1.31 AV3-4 Augustana Upper -85.1683 -174.1513 1399 72 148 0.85 Continued

Table B2 continued

AV3-5 Augustana Upper -85.1676 -174.1528 1370 72 73 1.10 AV3-6 Augustana Upper -85.1672 -174.1519 1356 72 33 0.71 AV3-7 Augustana Upper -85.1671 -174.1508 1353 72 20 2.54 AV3-8 Augustana Upper -85.1674 -174.1452 1351 72 35 9.38 BP1-1 Bennett Upper -85.2126 -177.3386 1395 82 1150 0.69 BP1-2 Bennett Upper -85.2105 -177.3285 1365 82 1030 0.64 BP1-3 Bennett Upper -85.2078 -177.3224 1352 82 887 0.67 BP1-4 Bennett Upper -85.2050 -177.3292 1329 82 637 0.62 BP1-5 Bennett Upper -85.2032 -177.3433 1297 82 433 0.60 BP1-6 Bennett Upper -85.2015 -177.3620 1251 82 187 0.52

169 BP1-7 Bennett Upper -85.2009 -177.3721 1232 82 73 0.78

BP1-8 Bennett Upper -85.2006 -177.3758 1225 82 24 1.08 BP2-1 Bennett Upper -85.2121 -177.3576 1410 82 1007 0.66 BP2-2 Bennett Upper -85.2116 -177.3589 1405 82 950 1.42 BP2-3 Bennett Upper -85.2112 -177.3620 1386 82 898 2.92 BP2-4 Bennett Upper -85.2102 -177.3673 1378 82 782 1.67 BP2-5A* Bennett Upper -85.2072 -177.3887 1294 82 396 0.98 BP2-5B* Bennett Upper -85.2059 -177.4018 1269 82 225 1.07 BP2-7 Bennett Upper -85.2036 -177.3920 1235 82 93 0.89 BP2-8 Bennett Upper -85.2024 -177.3907 1222 82 27 1.12 BP3-1 Bennett Upper -85.2144 -177.4048 1422 82 1041 1.43 BP3-2 Bennett Upper -85.2135 -177.4086 1400 82 935 4.00 Continued

Table B2 continued

BP3-3 Bennett Upper -85.2126 -177.4100 1371 82 831 1.55 BP3-4 Bennett Upper -85.2116 -177.4174 1351 82 713 2.30 BP3-5 Bennett Upper -85.2100 -177.4193 1329 82 545 1.59 BP3-6 Bennett Upper -85.2062 -177.4167 1277 82 173 1.32 BP3-7 Bennett Upper -85.2053 -177.4149 1246 82 103 0.99 BP3-8 Bennett Upper -85.2049 -177.4136 1238 82 75 6.38 MF1-1 Franke Lower -84.6251 -176.7705 409 9 637 1.25 MF1-2 Franke Lower -84.6262 -176.7664 620 9 537 5.00 MF1-3 Franke Lower -84.6264 -176.7611 551 9 467 3.53 MF1-4 Franke Lower -84.6265 -176.7549 555 9 400 3.52

170 MF1-5 Franke Lower -84.6265 -176.7498 531 9 350 0.53

MF1-6 Franke Lower -84.6265 -176.7434 485 9 283 1.03 MF2-1 Franke Lower -84.6236 -176.7353 480 9 323 3.15 MF2-2 Franke Lower -84.6235 -176.7328 466 9 300 1.49 MF2-4 Franke Lower -84.6237 -176.7252 424 9 218 0.39 MF3-1 Franke Lower -84.6227 -176.7507 526 9 502 1.71 MF3-2 Franke Lower -84.6227 -176.7456 520 9 454 2.33 MFX-1 Franke Lower -84.6241 -176.7076 332 9 33 7.71 MFX-2 Franke Lower -84.6241 -176.7085 329 9 34 4.85 MFX-3 Franke Lower -84.6241 -176.7096 333 9 56 8.14 MH1-1 Heekin Middle -85.0466 -177.2275 1661 63 1942 3.75 MH1-2 Heekin Middle -85.0441 -177.2334 1393 63 1723 0.97 Continued

Table B2 continued

MH1-3 Heekin Middle -85.0419 -177.2421 1196 63 1540 1.93 MH1-4 Heekin Middle -85.0404 -177.2420 1170 63 1386 1.54 MH1-5 Heekin Middle -85.0381 -177.2370 1089 63 1157 1.50 MH1-6 Heekin Middle -85.0339 -177.2206 1058 63 699 0.75 MH1-7 Heekin Middle -85.0310 -177.1790 1058 63 204 1.52 MH1-8 Heekin Middle -85.0301 -177.1751 1029 63 86 4.74 MH2-5 Heekin Middle -85.0467 -177.4162 990 63 654 0.77 MH2-6 Heekin Middle -85.0485 -177.4213 1011 63 458 1.34 MH2-7 Heekin Middle -85.0504 -177.4161 965 63 263 0.74 MH2-8 Heekin Middle -85.0528 -177.4099 1135 63 17 0.84

171 MH3-1 Heekin Middle -85.0332 -177.3292 1200 63 1098 2.89

MH3-10 Heekin Middle -85.0286 -177.1937 1030 63 39 10.79 MH3-2 Heekin Middle -85.0329 -177.3167 1175 63 902 2.04 MH3-3 Heekin Middle -85.0329 -177.3021 1140 63 451 1.76 MH3-4 Heekin Middle -85.0308 -177.2549 1090 63 450 1.50 MH3-5 Heekin Middle -85.0311 -177.2489 1080 63 507 0.97 MH3-6 Heekin Middle -85.0316 -177.2427 1064 63 510 1.44 MH3-7 Heekin Middle -85.0320 -177.2375 1050 63 533 0.70 MH3-8 Heekin Middle -85.0324 -177.2265 1045 63 127 2.91 MH3-9 Heekin Middle -85.0295 -177.1923 1056 63 991 0.94 MSP10X Speed Lower -84.4649 -177.1615 313 1 10 6.13 MSP1-1 Speed Lower -84.4830 -176.5128 316 1 500 0.11 Continued

Table B2 continued

MSP1-2 Speed Lower -84.4831 -176.5035 264 1 408 0.16 MSP1-3 Speed Lower -84.4830 -176.4973 237 1 340 0.98 MSP13X Speed Lower -84.4649 -177.1277 316 1 1 4.24 MSP1-4 Speed Lower -84.4830 -176.4805 188 1 167 0.43 MSP2-1 Speed Lower -84.4819 -176.5070 270 1 407 0.14 MSP2-2 Speed Lower -84.4819 -176.4975 225 1 311 0.16 MSP2-3 Speed Lower -84.4815 -176.4947 232 1 281 4.66 MSP2-4 Speed Lower -84.4811 -176.4864 181 1 186 0.21 MSP3-1 Speed Lower -84.4819 -176.4879 150 1 214 1.23 MSP3-2 Speed Lower -84.4809 -176.4793 163 1 99 0.65

172 MSP3-3 Speed Lower -84.4805 -176.4827 168 1 117 3.97

MSP3-4 Speed Lower -84.4802 -176.4881 193 1 161 0.19 MSP4-1 Speed Lower -84.4661 -177.1224 276 1 15 0.28 MSP4-2A Speed Lower -84.4657 -177.1357 287 1 11 0.33 MSP4-2B Speed Lower -84.4657 -177.1357 287 1 11 0.60 MSP4-2C Speed Lower -84.4657 -177.1357 287 1 11 1.05 MSP4-4 Speed Lower -84.4647 -177.1685 308 1 4 5.43 MSPX-1 Speed Lower -84.4784 -176.4787 163 1 100 2.98 MSPX-2 Speed Lower -84.4815 -176.4787 163 1 111 0.29 MSPX-3 Speed Lower -84.4815 -176.4787 163 1 98 8.47 MSPX-4 Speed Lower -84.4816 -176.4766 167 1 86 2.88 MSPX-5 Speed Lower -84.4815 -176.4772 171 1 99 2.24 Continued

Table B2 continued

MSPX-6 Speed Lower -84.4808 -176.4818 175 1 124 3.63 MW 2-1 Wasko Lower -84.5596 -176.8112 321 10 2 12.27

MW 4-1 Wasko Lower 345 10 7 0.49

MW 4-2 Wasko Lower 352 10 1 1.31

MW 4-3A Wasko Lower 350 10 1 1.79

MW 4-3B Wasko Lower 350 10 1 1.68 NP1-1 Nilsen Lower -84.5346 -175.4222 611 1 12 2.73 NP1-2 Nilsen Lower -84.5341 -175.4055 608 1 30 0.27 NP1-3 Nilsen Lower -84.5339 -175.4026 593 1 52 0.57 NP2-1 Nilsen Lower -84.5348 -175.4262 688 1 10 2.46

173 NP2-2 Nilsen Lower -84.5351 -175.4271 674 1 15 0.76

NP2-3 Nilsen Lower -84.5353 -175.4278 670 1 43 0.40 NP2-4 Nilsen Lower -84.5357 -175.4303 668 1 37 4.44 NP3-1 Nilsen Lower -84.5346 -175.4232 683 1 66 3.99 NP3-2 Nilsen Lower -84.5337 -175.4019 645 1 307 1.80 NP3-3 Nilsen Lower -84.5341 -175.4076 655 1 283 2.48 NP3-4 Nilsen Lower -84.5344 -175.4166 673 1 146 4.86 NP4-1 Nilsen Lower -84.5317 -175.3360 621 1 10 0.06 NP4-3 Nilsen Lower -84.5319 -175.3350 648 1 8 0.05 NP4-5 Nilsen Lower -84.5332 -175.3152 685 1 1 1.20 NP4-6 Nilsen Lower -84.6227 -176.7501 557 1 495 0.47 RM1-1 Roberts Upper -85.4909 -177.1772 1801 120 1435 0.84 Continued

Table B2 continued

RM1-2 Roberts Upper -85.4905 -177.1703 1797 120 1345 1.08 RM1-3 Roberts Upper -85.4900 -177.1712 1771 120 1322 0.63 RM1-4 Roberts Upper -85.4894 -177.1664 1769 120 1238 0.58 RM1-5 Roberts Upper -85.4884 -177.1558 1745 120 1107 0.50 RM1-6 Roberts Upper -85.4876 -177.1315 1738 120 900 0.74 RM1-7 Roberts Upper -85.4873 -177.1112 1742 120 755 0.72 RM1-8 Roberts Upper -85.4868 -177.0994 1731 120 653 0.69 RM2-1 Roberts Upper -85.4879 -177.1844 1776 120 882 0.86 RM2-2 Roberts Upper -85.4878 -177.1834 1777 120 871 0.88 RM2-3 Roberts Upper -85.4877 -177.1794 1763 120 848 0.69

174 RM2-4 Roberts Upper -85.4873 -177.1753 1760 120 795 0.61

RM2-5 Roberts Upper -85.4868 -177.1639 1754 120 706 0.63 RM2-6 Roberts Upper -85.4864 -177.1619 1753 120 665 0.72 RM2-7 Roberts Upper -85.4861 -177.1586 1749 120 624 1.27 RM2-8 Roberts Upper -85.4857 -177.1549 1747 120 564 0.93 RM3-1 Roberts Upper -85.4869 -177.2072 1708 120 899 0.59 RM3-2 Roberts Upper -85.4866 -177.1980 1783 120 818 1.12 RM3-3 Roberts Upper -85.4861 -177.1921 1773 120 746 0.69 RM3-4 Roberts Upper -85.4845 -177.1816 1747 120 561 0.85 RM3-5 Roberts Upper -85.4830 -177.1608 1757 120 318 0.88 RM3-6 Roberts Upper -85.4822 -177.1525 1722 120 201 1.22 RM3-7 Roberts Upper -85.4818 -177.1338 1688 120 116 1.22 Continued

Table B2 continued

RM3-8 Roberts Upper -85.4812 -177.1326 1688 120 45 1.39 SH1-1 Schroeder Upper -85.3566 -175.1108 2221 94 1027 2.72 SH1-3 Schroeder Upper -85.3538 -175.1415 2098 94 1079 2.26 SH1-4 Schroeder Upper -85.3529 -175.1430 2091 94 1047 1.93 SH1-5 Schroeder Upper -85.3514 -175.1677 2045 94 1211 2.02 SH1-6 Schroeder Upper -85.3512 -175.1727 2039 94 1229 2.06 SH2-1 Schroeder Upper -85.3591 -175.0787 2119 94 931 1.79 SH2-2 Schroeder Upper -85.3591 -175.0868 2131 94 992 1.56 SH2-3 Schroeder Upper -85.3588 -175.0860 2127 94 969 2.51 SH2-4 Schroeder Upper -85.3585 -175.0890 2138 94 968 1.49

175 SH2-5 Schroeder Upper -85.3583 -175.0875 2131 94 947 1.85

SH2-6 Schroeder Upper -85.3581 -175.0870 2143 94 934 2.24 SH2-7 Schroeder Upper -85.3579 -175.0880 2131 94 915 3.68 SH2-8 Schroeder Upper -85.3577 -175.0867 2143 94 905 1.26 SH3-2 Schroeder Upper -85.3597 -175.0693 2137 94 900 3.01 SH3-3 Schroeder Upper -85.3603 -175.0792 2127 94 1006 3.99 SH3-4 Schroeder Upper -85.3599 -175.1039 2122 94 1142 6.07 SH3-5 Schroeder Upper -85.3588 -175.1198 2092 94 1214 3.03 SH3-6 Schroeder Upper -85.3578 -175.1449 2074 94 1310 3.59 SH3-7 Schroeder Upper -85.3573 -175.1522 2061 94 1322 2.63 SH3-8 Schroeder Upper -85.3569 -175.1621 2057 94 1381 1.69 TGV1-1 Thanksgiving Middle -84.9203 -176.9902 1298 45 1266 2.23 Continued

Table B2 continued

TGV1-2 Thanksgiving Middle -84.9199 -176.9831 1198 45 1186 2.07 TGV1-3 Thanksgiving Middle -84.9196 -176.9732 1189 45 1079 1.53 TGV1-4 Thanksgiving Middle -84.9187 -176.9616 1142 45 937 0.88 TGV1-5 Thanksgiving Middle -84.9176 -176.9317 1077 45 628 0.95 TGV1-6 Thanksgiving Middle -84.9167 -176.9121 1013 45 410 0.99 TGV1-7 Thanksgiving Middle -84.9161 -176.8989 982 45 267 0.76 TGV1-8 Thanksgiving Middle -84.9154 -176.8791 912 45 63 1.27 TGV2-1 Thanksgiving Middle -84.9190 -177.0603 1107 45 1879 0.96 TGV2-2 Thanksgiving Middle -84.9178 -177.0528 1091 45 1758 1.64 TGV2-3 Thanksgiving Middle -84.9164 -177.0218 1082 45 1427 1.06

176 TGV2-4 Thanksgiving Middle -84.9155 -176.9956 1086 45 1167 1.56

TGV2-5 Thanksgiving Middle -84.9145 -176.9688 1082 45 886 1.30 TGV2-6 Thanksgiving Middle -84.9150 -176.9069 990 45 311 0.55 TGV2-7 Thanksgiving Middle -84.9146 -176.8917 934 45 153 1.33 TGV2-8 Thanksgiving Middle -84.9145 -176.8860 912 45 105 1.88 TGV2- Thanksgiving Middle -84.9174 -177.0490 1080 45 1701 10.54 EXTRA TGV3-1 Thanksgiving Middle -84.9096 -177.0217 1479 45 1490 2.00 TGV3-2 Thanksgiving Middle -84.9108 -177.0011 1200 45 1266 2.34 TGV3-3 Thanksgiving Middle -84.9135 -176.9651 1079 45 870 0.97 TGV3-4 Thanksgiving Middle -84.9146 -176.9323 1056 45 546 1.00 TGV3-5 Thanksgiving Middle -84.9150 -176.9118 1008 45 357 0.82 Continued

Table B2 continued

TGV3-6 Thanksgiving Middle -84.9145 -176.9008 972 45 240 0.93 TGV3-7 Thanksgiving Middle -84.9141 -176.8960 956 45 191 1.27 TGV3-8 Thanksgiving Middle -84.9136 -176.8835 912 45 74 0.96 TGV3-9 Thanksgiving Middle -84.9137 -176.8832 911 45 74 7.02 TN1-1 Taylor Middle -84.9257 -176.0692 1137 45 168 2.58 TN1-2 Taylor Middle -84.9254 -176.0765 1093 45 199 0.62 TN1-3 Taylor Middle -84.9252 -176.0818 1061 45 139 3.82 TN1-4 Taylor Middle -84.9247 -176.0901 1051 45 56 2.91 TN1-5 Taylor Middle -84.9210 -176.0976 994 45 66 1.74 TN1-6 Taylor Middle -84.9187 -176.1080 955 45 48 3.00

177 TN1-7 Taylor Middle -84.9160 -176.1339 920 45 32 3.19

TN1-8 Taylor Middle -84.9148 -176.1509 873 45 5 1.09 TN1-9 Taylor Middle -84.9156 -176.1466 883 45 20 10.87 TN2-1 Taylor Middle -84.9238 -176.0988 1030 45 40 7.89 TN2-2 Taylor Middle -84.9242 -176.0996 1034 45 76 6.43 TN2-3 Taylor Middle -84.9245 -176.0998 1041 45 96 2.90 TN2-4 Taylor Middle -84.9255 -176.1047 1053 45 208 2.04 TN2-5 Taylor Middle -84.9264 -176.1060 1056 45 298 2.74 TN2-6 Taylor Middle -84.9266 -176.1063 1068 45 320 3.57 TN2-7 Taylor Middle -84.9266 -176.1094 1068 45 343 2.90 TN2-8 Taylor Middle -84.9266 -176.1108 1070 45 350 7.15 TN3-1 Taylor Middle -84.9227 -176.1242 1097 45 154 1.71 Continued

Table B2 continued

TN3-2 Taylor Middle -84.9217 -176.1252 1086 45 188 3.35 TN3-3 Taylor Middle -84.9215 -176.1366 1023 45 292 2.57 TN3-5 Taylor Middle -84.9182 -176.1282 940 45 193 2.07 Continued

178

Table B2 continued - -1 - -1 - -1 2- -1 3- -1 - -1 - -1 2+ -1 F (µg g ) Cl (µg g ) NO3 (µg g ) SO4 (µg g ) PO4 (µg kg ) ClO4 (µg kg ) ClO3 (µg kg ) Ca (µg g )

3.34 369 6881 1443 432.5 514

2.14 124 666 13.27 130 18 50.17

2.39 537 4523 198 515 26.5 748

1.90 42.17 184 17.61 43.5 61.50

2.55 167 765 4099 113.5 4.5 1078

2.94 534 2816 236 457 55 684

1.99 66.87 275 20.20 52.5 43 61.75

2.20 300 2090 2189 173 65 463

6.60 161 777 7501 82 44 2094

4.36 215 1593 474 150.5 40 1738 179 5.89 236 1925 435 176.5 1592

6.40 184 956 7603 110.5 4.5 2211

9.19 350 1345 4477 199 38 2089

6.14 302 1381 499 146 1694

2.25 8.48 14.67 2.76 3.6 0.77

3.39 17.80 56.14 52.36 18 1.5 9.50

4.94 293 1392 996 176.5 8 312

2.04 42.68 173 9.49 32.4 40.75

7.95 519 4284 553 545 55.5 2068

2.79 25.59 53.78 129 17.35 357

3.00 19.49 195 858 13.95 142 Continued

Table B2 continued

2.67 72.38 205 70.97 98.5 171

1.90 30.60 184 27.90 11.3 60.80

4.79 15.32 23.32 84.67 8.6 8.60

3.89 708 3240 2771 206 34.2 484 3.68 182 941 2815 85.18 81.5 18.5 759

0.39 318 911 115 118.5

1.95 315 862 275 161 46.37

4.33 5.64 1284 0.55

5.38 16.55 4.61 367

2.99 9.71 21.49 60.53 660 274 0.67

180 22.29 24.58 15.16 270 206 4.17 15.20

5.75 186 284 1556 44 351

8.92 822 3005 9547 249 2971

15.68 245 4719 9460 86 26.31 2408 1057 8679 116 428.5 65 4053

5.35 790 4.51 4087 41.5 48.5

3.33 6.75 707 99 11 1487

5.02 4.13 50.19 13.25 825 21.39 17.95 14.25 64.70 555 2.5 7.5 3.85 16.18 960 2374 9190 231 103 80 2332

5.41 343 1053 780 79.5 2630 24.20 1751 13838 8618 276 712.5 80 3001 Continued

Table B2 continued

22.22 272 1483 11602 114 70.5 4109 16.24 1098 3167 5585 273 241 70 4404

4.80 118 134 1250 25 199

1.47 133 92.51 106 13 67.42

3.45 135 3.52 107 2.79 15.35

4.03 5.37 0.79 895 2.315 4.90

3.66 4.70 701 3.45

3.46 9.47 3.42 2.12 620 5.15

3.06 9.38 1.41 0.86 510 3.23

3.65 0.32 4.78 245 2.80

181 1.76 3.79 625 3.80

1.30 4.53 17.82 100 1 3.94

2.25 5.63 1.43 0.08 178 6.58

1.00 10.18 8.06 0.35 1.70

1.26 3.24 21.75 392 3.65

2.72 6.43 1.84 0.72 524 4.25

1.33 5.46 6.22 138 6.90

3.27 6.88 1.37 522 7.15

3.10 7.81 1.35 6.10 43.59 3170 38435 15886 112 312.5 20 3384

2.65 35.21 55.69 4101 10.25 1317

13.95 1904 1691 8594 95.73 470 2801 Continued

Table B2 continued

9.51 269 397 274 64.5 92.23

11.58 33.55 40.80 10166 9.25 3180

391 430 534 123 148

8.33 435 471 201 100 102.5 54.15

200 80.31 39.06 34.55 12.00

12.11 28.46 34.95 855 6.6 25 139

11.96 40.08 33.90 399 595 6.55 79.52

20.18 13.39 37.23 20.75 706 5.41

2.86 67.03 61.13 116 13.95 24.05 14.60 889 1400 13637 76.09 115 9 2865

182 15.84 4827 4841 7670 109 21.1 1954

4.22 694 1842 5434 665 1702

7.48 1063 2998 3816 171 2131

5.10 1406 1905 308 675 387 13.12 1090 1077 5706 124 195 6 1929 4.07 19.84 8.79 156 399 342 12.5 15.76 18.50 6060 6601 15751 166 3.455 4.5 3745

4.29 28.68 23.05 25.43 140 36.5 15.30

7.31 144 41.22 48.30 1589 8.2 3.89

4.44 10.43 2644 87.60

2.65 3.36 0.91 2.06

6.03 0.31 1.97 Continued

Table B2 continued

5.39 0.37 2.96

2.56 3.16 266 39.52

0.96 10.85 0.44 52.5 4.31

3.42 3.37

7.69 0.69 138 3.28

1.74 5.89 1.16 0.97 2.91

2.28 2.10 5.00 95.08 4.48

2.84 4.70 0.10 3.915 5.00

2.13 6.58 6.28 0.90 105 6.80

2.18 6.13 1.05 0.26 149 7.30

183 0.91 12.59 7.55 101 5.55

2.08 3.24 39.14

0.93 15.24 0.49 82.75 1 76.57

27.14 0.79 4195 6.50

3.37 7.21 3.60 1.79 1052 99.36

1.89 7.08 1.62 0.36 135 0.5 31.88

0.70 3.69 1.22 82.61 5.40

1.21 4.83 1.65 516 3.83

3.64 0.46 3.32

2.15 6.10 3.54 1.85 18.50

0.74 3.02 5.80 5.30

1.72 5.99 0.96 0.68 4.40 Continued

Table B2 continued

3.10 3.49 5.47 4.04

2.16 8.39 541 4.61

0.89 4.31 0.54 7.20

2.14 5.60 5.35 0.22 181 6.90

2.06 6.20 0.96 0.24

1.09 5.46 80.93 224 29.00

5.57 26.41 61.15 1459 6.24

3.05 8.86 1.19 4.95 1058 15.85

4.95 9.60 11.20

2.69 6.29 1.31 1.43 541 6.20

184 2.45 23.19 5.28 35.01 90.40

7.39 8.47 414 6.82

3.02 1.54 9.13 6.33

0.98 12.38 5.00 223 50.10

2.31 9.66 5.28 3.11 96.82 21.85

1.08 12.49 8.61 3.5 1 17.72

2.73 4.51 1.15 4.37

5.16 1.21 4.39

2.69 4.81 6.20

2.51 5.66 0.80 7.88

9.51 4.84 49.12 7902 430 2200 1098 33.47 11.58 426 28202 166 252.5 5300 2745 Continued

Table B2 continued

5.47 11.55 84.79 65.21 64.5 4500 37.96

6.56 2.92 9.09 398 66.5 1100 17.40

9.01 3.22 30.31 556 76 1000 64.25

8.05 10.29 29.24 393 40.85 1650 1284

4.83 10.26 75.96 54.20 555 5250 63.34

7.25 1.59 12.66 290 44.55 1250 8.85 31.02 139 1455 39909 125 240 5700 6.60

16.45 3.67 289 8647 755 9950 1357

9.46 6.64 169 207 64 1500 381

9.14 5.79 11.91 4.67 9.3 80 1.44

185 10.86 3.62 351 3419 205 1650 552

9.31 15.73 176 258 197 850 630

6.30 5.92 5.43 58.88 5.85 70.5 78.75

37.69 4.69 1012 13323 1000 4200 1191

7.39 23.08 128 2391 84.5 1650 428 23.96 79.13 3864 32733 322 481.5 14500 1947

8.23 10.84 57.14 188 19.95 2325 538 23.75 7.13 452 14320 120 198 7450 1492

28.39 27.79 1.13 11760 94.46 657 13.45 11.37 432 14634 120 117 5850 2686 10.34 14.45 93.60 8386 97.81 23.2 850 2252 7.74 86.25 1.15 8132 104 29.85 80 2505 Continued

Table B2 continued

16.75 23.71 1168 16260 137 50 2553 16.52 51.92 2511 14187 134 2800 65 2280 3.04 86.48 3789 8456 97.69 4025 48.5 1992

5.67 78.55 3211 3564 2500 50 1717

2.41 48.11 1954 3835 1990 50 756

7.30 9.98 779 9767 1780 1869

7.74 7.93 2247 6579 4500 165 1458 10.51 14.34 39.89 11345 91.75 107.5 65 2803 24.23 11.12 5910 15892 103 4175 100 1590 7.58 6.11 1493 8590 83.06 6750 6 2244

186 15.15 13.10 1531 21698 123 3175 35 2016

19.42 17.17 1502 14328 1950 240 2277

11.20 4.65 979 11868 78.28 720 2506

9.15 8.00 7551 21009 190 3550 2674 120 14.01 12967 310 393 75000 6750 1913 62.74 9.67 1942 55266 288 5400 3100 1717 23.46 13.94 7921 42614 314 180 700 2828 33.57 9.79 1721 28321 173 915 550 1934

19.33 27.97 7066 23318 184 910

7.05 29.97 5286 12125 99.27 155 2685

29.40 2858 3805 10970 818 2827

16.10 3888 6602 263 3975 1527 Continued

Table B2 continued

30.85 4443 4644 4962 1480 1409

14.90 758 1859 2159 300 339

10.55 320 1328 769 150.5 104

2.44 22.77 37.53 1.89 9.8 2.90

5.96 21.46 19.98 37.95 324 8.2 2.46

2.84 13.59 12.53 4.18 4.89

3.56 42.56 36.05 77.66 135 5.38

15.80 881 1509 327 400 1164

3.51 104 130 10.05 169 36.55 14.00

19.62 2456 5375 49.29 77.65 855 527

187 10.22 1738 709 3350 24 5.5 1234

3.70 59.50 497 124 82.60 22.1 12.17

2.15 7.25 1.41 0.88 0.68

3.63 5.83 4.84 2.5 0.78

6.55 429 418 119 137 62.50

8.97 45.59 595 4448 26.6 1203

16.18 675 3136 11186 400 3098

6.80 303 362 86.81 95 155

18.18 864 986 245 254 659

6.87 956 1322 910 427.5 298

9.26 567 810 24.62 176.5 111

2.42 69.64 80.05 4.32 26.05 1.30 Continued

Table B2 continued

4.09 22.12 25.22 7.85 2.66

4.16 68.96 37.30 12.72 190 6.35 8.30 5932 6680 4988 100 1410 270 3801

1.38 17.33 6.15 27.72 94.10

3.11 12.89 3.85 6.36 218 48.86 5.42 13555 8813 2386 82.47 1490 250 2598

3.85 2125 1376 226 263 638

9.59 9311 4779 8320 880 300 4316

9.32 10272 6381 6438 1240 185 3726

6.14 119 42.04 420 936 12.85 36.10

188 4.04 7.60 10.40 65.24 109 6.2 65 35.85

3.57 198 106 63.86 23 29.50

0.84 11.70 2.34 9.81

3.15 5477 5319 9383 660 80 3394

7.72 10683 7406 4940 141 885 3234

4.91 4014 55.76 1302 525 6.5 1360

2.05 829 521 114 113 5 265

6.68 5934 8363 941 680 4157

3.81 168 70.36 990 24 5.5 322

6.85 41.93 56.50 935 25.95 354 12.84 1196 6.23 521 89.02 4.27 250 26.45

4.99 3942 4673 252 955 90 1662 Continued

Table B2 continued

5.99 103 48.94 68.91 468 18.45 35.75 Continued

189

Table B2 continued 2+ -1 + -1 + -1 2+ -1 -1 -1 -1 Mg (µg g ) Na (µg g ) K (µg g ) Sr (µg g ) H4SiO4 (µg g ) NH3 (µg kg ) N:P ratio (molar) Total salt (µg g )

511 1913 35.40 2.50 12.66 420 11672

85.68 219 10.31 0.38 17.73 54.63 1172

660 833 23.74 2.37 15.89 279 7529

27.10 48.96 5.65 0.45 8.32 389

303 349 23.15 2.50 26.34 87.77 6788

486 737 18.42 1.70 13.28 210 5516

37.40 77.53 5.45 0.35 8.64 277 547

321 597 20.70 1.25 19.46 47.81 5984

275 435 20.75 5.90 27.12 212 11276

382 323 24.34 5.07 15.51 178 4759 190 366 395 21.85 13.60 10.17 148 4989

364 398 20.09 2.03 27.53 78.39 11745

489 474 41.50 3.90 23.60 49.44 9278

396 363 24.30 2.40 24.79 213 4667

0.68 25.56 0.95 18.47 74.09 56.12

6.20 49.40 3.50 13.56 74.88 198

234 294 14.30 1.25 20.89 38.33 3541

20.95 56.94 5.45 10.42 37.46 351

750 839 35.29 3.64 31.49 901 9060

79.31 122 8.89 0.72 17.18 43.18 779

39.45 193 7.50 0.45 18.96 1458 Continued

Table B2 continued

70.32 196 9.33 0.76 19.62 56.46 798

19.30 77.38 7.22 15.69 31.35 410

5.30 60.60 3.95 21.71 265 207 35.95 1989 5.60 0.40 30.98 56.74 24103 9239

92.24 449 14.64 37.19 16926 5242

13.82 391 11.96 92.59 1750

21.61 760 16.36 2282

0.13 16.19 31.83 59.45 28.13

16.44 13.84 33.48 68.98 43.34

0.42 62.92 0.86 25.93 508 49.86 161

191 4.45 237 3.80 35.85 277 113 593

83.55 401 5.20 0.85 21.74 34.29 2874

394 1343 15.65 9.75 26.04 176 18116

38.81 69.51 14440 897 1797 24.52 10.20 17.81 105 14005 18952

21.70 4887

213 552 12.10 1.35 33.82 114 2982

0.12 24.39 22.77 93.24 97.92

240 3.55 56.54 85.25 39.36 366 421 1610 30.80 4.60 37.41 123 15762 16938

275 728 13.70 8.80 32.77 185 5838 817 5883 37.83 8.54 14.76 168 76894 33981 Continued

Table B2 continued

150 869 18.21 3.69 34.59 106 19932 18529 698 1419 32.64 6.91 43.62 341 17759 16427

41.01 431 5.86 0.36 25.33 2184

20.11 97.77 5.47 20.55 33.31 524

4.41 39.22 1.41 24.82 149 310

2.85 10.45 6.75 34.75 221 1.35 36.03

2.15 15.45 5.60 44.20 89.83 35.71

3.05 13.70 6.50 45.65 67.88 8.46 47.50

1.74 9.97 3.96 36.45 78.70 4.25 34.12

2.00 5.65 4.70 29.60 135 2.00 24.14

192 2.45 8.65 5.25 27.45 111 26.33

2.03 5.37 2.65 19.22 73.28 272 37.75

1.84 8.29 3.25 28.22 12.31 29.53

1.08 1.48 2.04 12.18 131 25.54

1.73 7.70 4.11 30.25 74.34 84.91 43.83

1.90 9.10 5.15 38.82 44.53 5.38 32.63

3.25 11.20 6.60 29.90 72.91 68.94 41.10

3.95 12.30 6.55 40.78 291 4.03 41.99

3.50 8.25 6.15 41.58 220 36.26 1412 4032 130 38.65 31.86 383 525564 66532

124 81.75 21.59 60.22 5717 902 1074 43.35 8.45 35.78 824 27054 17032 Continued

Table B2 continued

66.94 197 7.64 0.62 16.51 42.86 1315

104 255 21.60 3.35 43.06 523 13816

83.10 286 7.06 0.39 14.60 47.35 1879

34.64 431 8.86 23.11 37.89 7195 1645

4.80 178 4.55 15.66 97.77 519

25.85 222 8.65 0.50 28.61 184 1327

28.04 103 8.25 29.27 220 87.32 705

1.58 90.96 27.53 80.72 190

7.15 93.60 5.40 16.51 31.78 378 215 521 57.40 5.62 41.84 139 28191 19605

193 853 41.50 3.11 17.24 147 39.72 20206

419 875 21.84 4.20 19.89 98.71 67906 10998

480 1005 29.79 6.11 30.17 93.40 11537

262 707 20.30 2.50 25.49 231 5003

339 869 23.51 3.61 22.89 67.22 11050

5.74 89.22 4.34 25.55 49.49 13344 305 546 4985 79.99 9.99 29.65 738 33.79 37797

4.15 39.30 4.50 20.18 61028 145

1.25 270 2.33 43.47 169 39.72 520

2.45 6.10 9.10 10.13 1908 123

0.91 1.17 1.78 13.94 173 12.84

0.88 1.50 1.89 5.30 54.13 12.59 Continued

Table B2 continued

0.97 1.39 2.04 3.75 55.20 13.12

0.76 4.09 2.13 10.08 287 52.49

1.53 1.89 2.40 6.38 98.94 22.37

1.54 2.12 3.01 8.99 167 13.47

1.39 1.74 2.19 5.41 37.81 7.68 17.12

1.70 3.53 2.80 21.66 20.70

1.58 1.14 2.66 3.44 80.10 33.86 19.33

2.70 4.40 8.70 8.99 976 28.45

3.05 4.33 6.40 3.63 134 91.72 36.57

3.20 4.43 6.65 6.40 467 10.78 31.34

194 2.50 2.50 8.90 4.18 683 114 40.60

0.64 2.44 2.99 3.95 238 50.52

1.44 5.14 7.86 9.54 382 8.98 108

4.25 4.30 4.20 88.41 0.29 47.17 1.96 10.92 7.20 0.09 5.94 152 5.25 137

0.76 0.92 2.98 5.63 293 18.35 47.62

2.30 1.90 5.75 3.46 166 22.69 21.05

1.74 17.00 5.70 21.33 4.91 36.49

1.08 1.83 4.86 3.76 642 15.19

8.20 6.45 7.90 20.41 1078 54.69

2.35 1.60 4.50 7.17 150 23.31

2.15 2.20 3.95 8.35 121 22.06 Continued

Table B2 continued

0.87 7.34 3.48 16.93 118 27.79

1.39 9.26 2.76 13.69 244 29.11

3.05 6.85 3.55 9.76 80.20 26.39

1.36 3.08 2.93 8.04 263 45.28 27.78

8.87 167 9.46

3.45 17.05 5.10 17.61 54.22 142

1.30 55.92 14.19 19.84 142 172

4.65 15.75 12.60 17.45 103 1.73 67.95

0.66 11.13 2.64 15.65 109 40.19

1.40 11.20 4.61 12.77 87.07 3.70 35.66

195 12.30 49.00 10.75 12.59 70.48 228

1.23 22.85 4.52 25.52 52.36 51.70

0.94 5.57 2.24 8.49 167 28.77

8.20 10.20 25.40 17.31 561 335

4.65 16.85 10.85 11.02 65.12 83.61 74.67

2.37 7.20 5.06 7.10 1229 54.54

0.16 0.39 3.64 39.75 13.32

0.14 0.39 1.14 123 11.30

0.53 1.71 1.26 3.31 166 17.19

0.40 1.45 1.67 2.96 222 20.38

53.60 1896 34.50 2.50 22.68 200 11050 579 11134 103 11.09 23.76 197 3926 43245 Continued

Table B2 continued

6.92 316 7.68 13.38 49.38 536

5.80 166 7.60 24.96 128 614

13.55 175 8.95 35.69 860

76.73 426 20.75 1.90 30.24 110 2251

14.11 214 10.20 28.02 140 447

2.20 132 4.90 25.69 134 460

0.85 9.35 24.01 344 17767 41552

103 1935 37.56 2.35 21.32 200 12392

57.65 525 13.90 0.65 17.51 61.30 1370

0.38 53.08 1.56 19.61 44.21 87.98

196 80.00 818 20.30 1.40 5256

59.74 591 14.57 0.96 18.88 55.38 1756

18.20 169 7.55 34.57 84.91 350

175 2096 59.95 4.85 34.74 189 17905

47.20 614 24.85 1.20 22.62 99.89 3664 424 11886 122 9.25 26.98 722 18393 51090

59.11 308 26.74 0.86 30.97 139 1197 102 4453 59.32 5.16 19.57 268 5756 20914

72.58 4131 38.15 3.74 32.59 18.34 16720 343 5182 59.00 8.70 19.93 111 5526 23370 157 715 30.98 5.04 38.75 165 1466 11665

133 555 40.60 3.75 43.83 16.84 11464 Continued

Table B2 continued

842 2557 139 34.55 23.65 1055 13103 23594 823 2663 133 34.34 26.14 1702 28669 22702 698 1166 78.25 16.35 20.40 2702 59419 16289

577 1076 69.30 23.55 8.54 5077 10325

254 1026 50.80 3.70 20.66 2606 7932

698 874 46.10 11.50 20.35 3190 14063

424 709 80.70 5.30 29.99 543 11523 158 1146 94.49 27.35 41.53 2336 666 15638 1277 2906 85.81 12.96 21.27 408 88038 27714 293 1007 80.05 10.65 22.95 960 27539 13738

197 2514 2642 65.74 22.24 15.95 1242 19096 30520

240 3063 83.77 24.62 16.35 410 21556 446 977 43.90 19.85 17.81 1142 19150 16856 3129 3836 56.11 14.40 7.57 1035 60788 38290 6281 25260 440 36.49 27.52 2845 50567 47417 1750 19301 395 31.89 22.76 538 10336 80482 5781 16958 268 46.61 27.03 1012 38620 76455 1800 7470 171 32.85 40.99 715 15230 41494

27.60 545 58864 30433 953 1955 124 32.90 27.88 1145 81574 23198

840 2374 68.40 16.30 37.30 166 23789

1605 3791 71.73 15.85 31.52 109 17784 Continued

Table B2 continued

1083 3226 54.13 10.50 19.70 101 19863

183 1144 18.79 2.15 17.76 95.27 6477

117 377 15.05 1.00 19.72 104 3040

2.95 31.05 3.65 14.10 18.85 105

2.92 48.82 2.22 15.40 95.04 94.35 142

3.04 16.30 2.32 16.63 20.35 55.52

3.14 78.75 2.83 19.16 155 250

440 445 33.00 4.45 22.01 97.30 4820

12.90 115 6.85 21.95 690 1184 397 747 1366 60.80 6.70 18.65 184 106043 10608

198 506 498 36.30 6.00 8.21 62.37 8088

7.76 95.83 3.91 0.07 42.94 203 9217 804

0.41 9.18 1.02 0.01 13.95 55.96 22.97

0.58 14.57 1.07 0.01 15.87 107 31.30

69.10 282 12.35 0.60 31.84 548 1400

280 354 11.70 17.40 7.47 84.12 6963

569 1222 45.74 30.52 22.70 288 19979

121 287 14.05 1.05 23.86 73.46 1336

257 790 26.25 2.05 17.82 110 3848

229 735 23.85 2.15 18.49 345 4484

110 403 14.85 0.95 17.77 32.93 2051

0.90 93.03 1.91 11.78 75.30 254 Continued

Table B2 continued

2.06 37.01 2.16 17.98 141 95.32

4.80 71.50 4.50 20.98 193 301 210 808 3665 46.52 15.83 38.24 383 101855 25948

14.95 39.03 4.63 16.95 50.51 205 2.08 38.26 4.90 0.08 21.87 73.15 27.03 121 1226 6340 74.60 25.36 25.05 170 163698 35026

166 817 18.20 6.40 18.23 34.81 5377

910 3678 79.55 20.25 36.41 185 31424

1067 3979 59.20 21.49 27.43 256 31953

7.85 338 10.70 49.26 163 68.79 981

199 7.00 60.50 4.65 36.38 104 146 195

8.75 181 5.45 30.32 78.43 596

1.13 15.87 1.50 12.10 72.66 43.20

752 2615 45.25 9.49 32.95 158 26998 1053 6086 36.93 16.90 14.64 164 80591 33465

469 1515 36.70 7.30 20.83 387 8765

96.80 318 14.50 1.05 41.35 77.76 2161

539 4605 36.17 4.96 28.29 258 24588

41.75 157 12.05 1.05 34.02 399 1766

18.20 113 18.55 0.75 60.78 40.74 1544

5.40 51.95 5.05 32.80 192 107 1825

405 2251 41.55 9.80 21.61 252 13243

6.45 168 6.40 40.63 177 160 444

Table B3. Random forest model imputed training dataset.

Sample ID Latitude Longitude Elevation (m) Distance from coast (km) Distance from glacier (m) Soil moisture (%) NP4-3 -84.5319 -175.3350 648 1 8 0.05 NP4-1 -84.5317 -175.3360 621 1 10 0.06 MSP1-1 -84.4830 -176.5128 316 1 500 0.11 MSP2-1 -84.4819 -176.5070 270 1 407 0.14 MSP2-2 -84.4819 -176.4975 225 1 311 0.16 MSP3-4 -84.4802 -176.4881 193 1 161 0.19 MSP2-4 -84.4811 -176.4864 181 1 186 0.21 NP1-2 -84.5341 -175.4055 608 1 30 0.27 MSP4-1 -84.4661 -177.1224 276 1 15 0.28

200 MSPX-2 -84.4815 -176.4787 163 1 111 0.29 MSP4-2A -84.4657 -177.1357 287 1 11 0.33 NP2-3 -84.5353 -175.4278 670 1 43 0.40 MSP1-4 -84.4830 -176.4805 188 1 167 0.43 NP4-6 -84.6227 -176.7501 557 1 495 0.47 MW 4-1 -84.5596 -176.8112 345 10 7 0.49 RM1-5 -85.4884 -177.1558 1745 120 1107 0.50 BP1-6 -85.2015 -177.3620 1251 82 187 0.52 AV3-2 -85.1703 -174.1475 1430 72 352 0.54 TGV2-6 -84.9150 -176.9069 990 45 311 0.55 NP1-3 -84.5339 -175.4026 593 1 52 0.57 Continued

Table B3 continued

RM1-4 -85.4894 -177.1664 1769 120 1238 0.58 RM3-1 -85.4869 -177.2072 1708 120 899 0.59 AV1-7 -85.1762 -174.1071 1451 72 1031 0.59 BP1-5 -85.2032 -177.3433 1297 82 433 0.60 RM2-4 -85.4873 -177.1753 1760 120 795 0.61 BP1-4 -85.2050 -177.3292 1329 82 637 0.62 AV1-4 -85.1762 -174.0975 1473 72 1071 0.62 TN1-2 -84.9254 -176.0765 1093 45 199 0.62 RM1-3 -85.4900 -177.1712 1771 120 1322 0.63 AV3-1 -85.1707 -174.1483 1439 72 392 0.63

201 AV2-7 -85.1680 -174.1389 1374 72 104 0.64

BP1-2 -85.2105 -177.3285 1365 82 1030 0.64 MSP3-2 -84.4809 -176.4793 163 1 99 0.65 BP2-1 -85.2121 -177.3576 1410 82 1007 0.66 BP1-3 -85.2078 -177.3224 1352 82 887 0.67 BP1-1 -85.2126 -177.3386 1395 82 1150 0.69 RM2-3 -85.4877 -177.1794 1763 120 848 0.69 RM3-3 -85.4861 -177.1921 1773 120 746 0.69 MH3-7 -85.0320 -177.2375 1050 63 533 0.70 AV3-6 -85.1672 -174.1519 1356 72 33 0.71 RM1-7 -85.4873 -177.1112 1742 120 755 0.72 RM2-6 -85.4864 -177.1619 1753 120 665 0.72 Continued

Table B3 continued

AV2-8 -85.1676 -174.1393 1378 72 61 0.73 MH2-7 -85.0504 -177.4161 965 63 263 0.74 RM1-6 -85.4876 -177.1315 1738 120 900 0.74 MH1-6 -85.0339 -177.2206 1058 63 699 0.75 NP2-2 -84.5351 -175.4271 674 1 15 0.76 MH2-5 -85.0467 -177.4162 990 63 654 0.77 BP1-7 -85.2009 -177.3721 1232 82 73 0.78 AV1-8 -85.1760 -174.1165 1433 72 993 0.80 TGV3-5 -84.9150 -176.9118 1008 45 357 0.82 AV1-1 -85.1762 -174.0885 1489 72 1108 0.82

202 MH2-8 -85.0528 -177.4099 1135 63 17 0.84

AV3-4 -85.1683 -174.1513 1399 72 148 0.85 RM3-4 -85.4845 -177.1816 1747 120 561 0.85 RM2-1 -85.4879 -177.1844 1776 120 882 0.86 RM3-5 -85.4830 -177.1608 1757 120 318 0.88 RM2-2 -85.4878 -177.1834 1777 120 871 0.88 TGV1-4 -84.9187 -176.9616 1142 45 937 0.88 AV1-5 -85.1761 -174.1005 1468 72 1049 0.90 RM2-8 -85.4857 -177.1549 1747 120 564 0.93 TGV3-6 -84.9145 -176.9008 972 45 240 0.93 MH3-9 -85.0295 -177.1923 1056 63 991 0.94 AV1-6 -85.1762 -174.1026 1455 72 1054 0.94 Continued

Table B3 continued

TGV1-5 -84.9176 -176.9317 1077 45 628 0.95 TGV2-1 -84.9190 -177.0603 1107 45 1879 0.96 TGV3-3 -84.9135 -176.9651 1079 45 870 0.97 MH3-5 -85.0311 -177.2489 1080 63 507 0.97 MH1-2 -85.0441 -177.2334 1393 63 1723 0.97 MSP1-3 -84.4830 -176.4973 237 1 340 0.98 AV2-3 -85.1701 -174.1349 1390 72 329 0.98 BP3-7 -85.2053 -177.4149 1246 82 103 0.99 TGV1-6 -84.9167 -176.9121 1013 45 410 0.99 TGV3-4 -84.9146 -176.9323 1056 45 546 1.00

203 AV1-3 -85.1763 -174.0943 1480 72 1086 1.02

MF1-6 -84.6265 -176.7434 485 9 283 1.03 AV2-1 -85.1706 -174.1338 1410 72 388 1.04 TGV2-3 -84.9164 -177.0218 1082 45 1427 1.06 BP2-5B -85.2059 -177.4018 1269 82 225 1.07 RM1-2 -85.4905 -177.1703 1797 120 1345 1.08 BP1-8 -85.2006 -177.3758 1225 82 24 1.08 TN1-8 -84.9148 -176.1509 873 45 5 1.09 AV3-5 -85.1676 -174.1528 1370 72 73 1.10 RM3-2 -85.4866 -177.1980 1783 120 818 1.12 BP2-8 -85.2024 -177.3907 1222 82 27 1.12 AV2-6 -85.1689 -174.1374 1376 72 195 1.14 Continued

Table B3 continued

NP4-5 -84.5332 -175.3152 685 1 1 1.20 RM3-7 -85.4818 -177.1338 1688 120 116 1.22 RM3-6 -85.4822 -177.1525 1722 120 201 1.22 MF1-1 -84.6251 -176.7705 409 9 637 1.25 SH2-8 -85.3577 -175.0867 2143 94 905 1.26 TGV3-7 -84.9141 -176.8960 956 45 191 1.27 RM2-7 -85.4861 -177.1586 1749 120 624 1.27 TGV1-8 -84.9154 -176.8791 912 45 63 1.27 TGV2-5 -84.9145 -176.9688 1082 45 886 1.30 AV3-3 -85.1694 -174.1462 1416 72 251 1.31

204 BP3-6 -85.2062 -177.4167 1277 82 173 1.32

TGV2-7 -84.9146 -176.8917 934 45 153 1.33 MH2-6 -85.0485 -177.4213 1011 63 458 1.34 RM3-8 -85.4812 -177.1326 1688 120 45 1.39 BP2-2 -85.2116 -177.3589 1405 82 950 1.42 MH3-6 -85.0316 -177.2427 1064 63 510 1.44 AV2-5 -85.1691 -174.1372 1388 72 226 1.46 SH2-4 -85.3585 -175.0890 2138 94 968 1.49 MF2-2 -84.6235 -176.7328 466 9 300 1.49 MH3-4 -85.0308 -177.2549 1090 63 450 1.50 MH1-5 -85.0381 -177.2370 1089 63 1157 1.50 TGV1-3 -84.9196 -176.9732 1189 45 1079 1.53 Continued

Table B3 continued

MH1-4 -85.0404 -177.2420 1170 63 1386 1.54 BP3-3 -85.2126 -177.4100 1371 82 831 1.55 SH2-2 -85.3591 -175.0868 2131 94 992 1.56 TGV2-4 -84.9155 -176.9956 1086 45 1167 1.56 BP3-5 -85.2100 -177.4193 1329 82 545 1.59 TGV2-2 -84.9178 -177.0528 1091 45 1758 1.64 BP2-4 -85.2102 -177.3673 1378 82 782 1.67 MW 4-3B -84.5596 -176.8112 350 10 1 1.68 SH3-8 -85.3569 -175.1621 2057 94 1381 1.69 TN3-1 -84.9227 -176.1242 1097 45 154 1.71

205 MF3-1 -84.6227 -176.7507 526 9 502 1.71

MH3-3 -85.0329 -177.3021 1140 63 451 1.76 SH2-1 -85.3591 -175.0787 2119 94 931 1.79 MW 4-3A -84.5596 -176.8112 350 10 1 1.79 NP3-2 -84.5337 -175.4019 645 1 307 1.80 SH2-5 -85.3583 -175.0875 2131 94 947 1.85 TGV2-8 -84.9145 -176.8860 912 45 105 1.88 SH1-4 -85.3529 -175.1430 2091 94 1047 1.93 TGV3-1 -84.9096 -177.0217 1479 45 1490 2.00 SH1-5 -85.3514 -175.1677 2045 94 1211 2.02 MH3-2 -85.0329 -177.3167 1175 63 902 2.04 TN2-4 -84.9255 -176.1047 1053 45 208 2.04 Continued

Table B3 continued

SH1-6 -85.3512 -175.1727 2039 94 1229 2.06 TN3-5 -84.9182 -176.1282 940 45 193 2.07 TGV1-1 -84.9203 -176.9902 1298 45 1266 2.23 MSPX-5 -84.4815 -176.4772 171 1 99 2.24 SH2-6 -85.3581 -175.0870 2143 94 934 2.24 SH1-3 -85.3538 -175.1415 2098 94 1079 2.26 BP3-4 -85.2116 -177.4174 1351 82 713 2.30 TGV3-2 -84.9108 -177.0011 1200 45 1266 2.34 NP2-1 -84.5348 -175.4262 688 1 10 2.46 NP3-3 -84.5341 -175.4076 655 1 283 2.48

206 SH2-3 -85.3588 -175.0860 2127 94 969 2.51

AV3-7 -85.1671 -174.1508 1353 72 20 2.54 TN3-3 -84.9215 -176.1366 1023 45 292 2.57 SH3-7 -85.3573 -175.1522 2061 94 1322 2.63 SH1-1 -85.3566 -175.1108 2221 94 1027 2.72 NP1-1 -84.5346 -175.4222 611 1 12 2.73 TN2-5 -84.9264 -176.1060 1056 45 298 2.74 MSPX-4 -84.4816 -176.4766 167 1 86 2.88 MH3-1 -85.0332 -177.3292 1200 63 1098 2.89 TN2-3 -84.9245 -176.0998 1041 45 96 2.90 TN1-4 -84.9247 -176.0901 1051 45 56 2.91 MH3-8 -85.0324 -177.2265 1045 63 127 2.91 Continued

Table B3 continued

BP2-3 -85.2112 -177.3620 1386 82 898 2.92 MSPX-1 -84.4784 -176.4787 163 1 100 2.98 TN1-6 -84.9187 -176.1080 955 45 48 3.00 SH3-5 -85.3588 -175.1198 2092 94 1214 3.03 MF2-1 -84.6236 -176.7353 480 9 323 3.15 TN1-7 -84.9160 -176.1339 920 45 32 3.19 TN3-2 -84.9217 -176.1252 1086 45 188 3.35 MF1-4 -84.6265 -176.7549 555 9 400 3.52 MF1-3 -84.6264 -176.7611 551 9 467 3.53 SH3-6 -85.3578 -175.1449 2074 94 1310 3.59

207 MSPX-6 -84.4808 -176.4818 175 1 124 3.63

SH2-7 -85.3579 -175.0880 2131 94 915 3.68 MH1-1 -85.0466 -177.2275 1661 63 1942 3.75 TN1-3 -84.9252 -176.0818 1061 45 139 3.82 MSP3-3 -84.4805 -176.4827 168 1 117 3.97 SH3-3 -85.3603 -175.0792 2127 94 1006 3.99 BP3-2 -85.2135 -177.4086 1400 82 935 4.00 MSP13X -84.4649 -177.1277 316 1 1 4.24 NP2-4 -84.5357 -175.4303 668 1 37 4.44 MSP2-3 -84.4815 -176.4947 232 1 281 4.66 MH1-8 -85.0301 -177.1751 1029 63 86 4.74 NP3-4 -84.5344 -175.4166 673 1 146 4.86 Continued

Table B3 continued

MF1-2 -84.6262 -176.7664 620 9 537 5.00 MSP4-4 -84.4647 -177.1685 308 1 4 5.43 SH3-4 -85.3599 -175.1039 2122 94 1142 6.07 MSP10X -84.4649 -177.1615 313 1 10 6.13 BP3-8 -85.2049 -177.4136 1238 82 75 6.38 TGV3-9 -84.9137 -176.8832 911 45 74 7.02 TN2-8 -84.9266 -176.1108 1070 45 350 7.15 MFX-1 -84.6241 -176.7076 332 9 33 7.71 TN2-1 -84.9238 -176.0988 1030 45 40 7.89 MFX-3 -84.6241 -176.7096 333 9 56 8.14

208 MSPX-3 -84.4815 -176.4787 163 1 98 8.47

TGV2- -84.9174 -177.0490 1080 45 1701 10.54 EXTRA MH3-10 -85.0286 -177.1937 1030 63 39 10.79 TN1-9 -84.9156 -176.1466 883 45 20 10.87 MW 2-1 -84.5596 -176.8112 321 10 2 12.27 Continued

Table B3 continued

- -1 - -1 - -1 2- -1 3- -1 - -1 - -1 F (µg g ) Cl (µg g ) NO3 (µg g ) SO4 (µg g ) PO4 (µg kg ) ClO4 (µg kg ) ClO3 (µg kg ) 0.00 5.16 1.21 0.00 0.00 0.00 0.00 2.73 4.51 1.15 0.00 0.00 0.00 0.00 2.65 3.36 0.91 0.00 0.00 0.00 0.00 0.00 3.42 0.00 0.00 0.00 0.00 0.00 0.00 7.69 0.69 0.00 138 0.00 0.00 0.91 12.59 7.55 0.00 101 0.00 0.00 0.00 2.28 2.10 5.00 95.08 0.00 0.00 5.57 26.41 0.00 61.15 1459 0.00 0.00 0.00 2.08 0.00 3.24 0.00 0.00 0.00 209 1.21 4.83 1.65 0.00 516 0.00 0.00

0.93 15.24 0.49 0.00 82.75 1 0.00 2.45 23.19 5.28 35.01 0.00 0.00 0.00 0.96 10.85 0.44 0.00 0.00 0.00 52.5 2.51 5.66 0.80 0.00 0.00 0.00 0.00 0.00 2.16 0.00 8.39 541 0.00 0.00 9.01 3.22 30.31 556 0.00 76 1000 0.00 5.38 16.55 4.61 367 0.00 0.00 2.04 42.68 173 9.49 0.00 32.4 0.00 3.70 59.50 497 124 82.60 22.1 0.00 3.05 8.86 1.19 4.95 1058 0.00 0.00 Continued

Table B3 continued

6.56 2.92 9.09 398 0.00 66.5 1100 7.39 23.08 128 2391 0.00 84.5 1650 1.99 66.87 275 20.20 0.00 52.5 43 4.33 5.64 0.00 0.00 1284 0.00 0.00 9.14 5.79 11.91 4.67 0.00 9.3 80 1.95 315 862 275 0.00 161 0 1.90 42.17 184 17.61 0.00 43.5 0.00 1.38 17.33 6.15 27.72 0.00 0.00 0.00 5.47 11.55 84.79 65.21 0.00 64.5 4500 4.94 293 1392 996 0.00 176.5 8

210 2.25 8.48 14.67 2.76 0.00 3.6 0.00

3.68 182 941 2815 85.18 81.5 18.5 2.13 6.58 6.28 0.90 105 0.00 0.00 5.75 186 284 1556 0.00 44 0 0.39 318 911 115 0.00 118.5 0 3.89 708 3240 2771 206 34.2 0 9.46 6.64 169 207 0.00 64 1500 8.23 10.84 57.14 188 0.00 19.95 2325 18.50 6060 6601 15751 166 3.455 4.5 2.67 72.38 205 70.97 0.00 98.5 0.00 4.83 10.26 75.96 54.20 0.00 555 5250 9.31 15.73 176 258 0.00 197 850 Continued

Table B3 continued

3.39 17.80 56.14 52.36 0.00 18 1.5 20.18 13.39 37.23 20.75 706 0.00 0.00 8.05 10.29 29.24 393 0.00 40.85 1650 0.00 391 430 534 0.00 123 0 2.69 6.29 1.31 1.43 541 0.00 0.00 12.11 28.46 34.95 855 0.00 6.6 25 2.99 9.71 21.49 60.53 660 274 0.00 2.20 300 2090 2189 0.00 173 65 6.87 956 1322 910 0.00 427.5 0 3.34 369 6881 1443 0.00 432.5 0

211 2.86 67.03 61.13 116 0.00 13.95 0.00

2.79 25.59 53.78 129 0.00 17.35 0.00 23.75 7.13 452 14320 120 198 7450 31.02 139 1455 39909 125 240 5700 28.39 27.79 1.13 11760 94.46 0.00 6650.00 16.45 3.67 289 8647 0.00 755 9950 14.90 758 1859 2159 0.00 300 0 2.55 167 765 4099 0.00 113.5 4.5 37.69 4.69 1012 13323 0.00 1000 4200 9.26 567 810 24.62 0.00 176.5 0 7.31 144 41.22 48.30 1589 8.2 0 2.94 534 2816 236 0.00 457 55 Continued

Table B3 continued

10.55 320 1328 769 0.00 150.5 0 3.56 42.56 36.05 77.66 0.00 135 0.00 6.80 303 362 86.81 0.00 95 0 13.12 1090 1077 5706 124 195 6 2.65 35.21 55.69 4101 0.00 10.25 0.00 0.00 5.39 0.37 0.00 0.00 0.00 0.00 5.89 236 1925 435 0.00 176.5 0.00 1.47 133 92.51 106 0.00 13 0.00 2.44 22.77 37.53 1.89 0.00 9.8 0.00 18.18 864 986 245 0.00 254 0.00

212 2.39 537 4523 198 0.00 515 26.5

1.76 3.79 0.00 0.00 625 0.00 0.00 6.60 161 777 7501 0.00 82 44 3.51 104 130 10.05 169 36.55 0.00 3.33 6.75 4.51 13.25 0.00 99 11 33.47 11.58 426 28202 166 252.5 5300 22.29 24.58 15.16 270 206 4.17 0 6.14 119 42.04 420 936 12.85 0 3.00 19.49 195 858 0.00 13.95 0 23.96 79.13 3864 32733 322 481.5 14500 21.39 17.95 14.25 64.70 555 2.5 7.5 6.14 302 1381 499 0.00 146 0 Continued

Table B3 continued

2.69 4.81 0.00 0.00 0.00 0.00 0.00 10.34 14.45 93.60 8386 97.81 23.2 850 13.45 11.37 432 14634 120 117 5850 4.03 5.37 0.79 0.00 895 2.315 0.00 11.20 4.65 979 11868 78.28 720 0 2.42 69.64 80.05 4.32 0.00 26.05 0.00 6.30 5.92 5.43 58.88 0.00 5.85 70.5 2.84 13.59 12.53 57.80 0.00 4.18 0.00 10.22 1738 709 3350 0.00 24 5.5 7.95 519 4284 553 0.00 545 55.5

213 4.80 118 134 1250 0.00 25 0

2.15 7.25 1.41 0.88 0.00 0.00 0.00 11.96 40.08 33.90 399 595 6.55 0 7.74 86.25 1.15 8132 104 29.85 80 8.92 822 3005 9547 0.00 249 0 4.07 19.84 8.79 156 399 342 12.5 9.19 350 1345 4477 0.00 199 38 24.23 11.12 5910 15892 103 4175 100 2.25 5.63 1.43 0.08 178 0.00 0.00 5.10 1406 1905 308 0.00 675 0 11.58 33.55 40.80 10166 0.00 9.25 0.00 30.85 4443 4644 4962 0.00 1480 0 Continued

Table B3 continued

9.51 269 397 274 0.00 64.5 0 24.20 1751 13838 8618 276 712.5 80 7.74 7.93 2247 6579 0.00 4500 165 19.62 2456 5375 49.29 77.65 855 0 16.24 1098 3167 5585 273 241 70 15.80 881 1509 327 0.00 400 0 26.31 2408 1057 8679 116 428.5 65 2.06 6.20 0.96 0.24 0.00 0.00 0.00 7.05 29.97 5286 12125 99.27 155 0 6.85 41.93 56.50 935 0.00 25.95 0.00

214 1.26 3.24 21.75 0.00 392 0.00 0.00

7.48 1063 2998 3816 0.00 171 0.00 7.30 9.98 779 9767 0.00 1780 100 2.14 5.60 5.35 0.22 181 0.00 0.00 0.98 12.38 5.00 223 0.00 0.00 0.00 7.58 6.11 1493 8590 83.06 6750 6 3.63 5.83 4.84 0.00 0.00 2.5 0.00 3.04 86.48 3789 8456 97.69 4025 48.5 8.97 45.59 595 4448 0.00 26.6 0.00 5.67 78.55 3211 3564 0.00 2500 50 4.22 694 1842 5434 0.00 665 0.00 7.72 10683 7406 4940 141 885 0 Continued

Table B3 continued

2.41 48.11 1954 3835 0.00 1990 50 5.99 103 48.94 68.91 468 18.45 0 29.40 2858 3805 10970 0.00 818 0 0.74 3.02 5.80 0.00 0.00 0.00 0.00 15.15 13.10 1531 21698 123 3175 35 16.52 51.92 2511 14187 134 2800 65 22.22 272 1483 11602 114 70.5 0 16.18 675 3136 11186 0.00 400 0 0.00 4.95 0.00 9.60 0.00 0.00 0.00 2.31 9.66 5.28 3.11 96.82 0.00 0.00

215 10.51 14.34 39.89 11345 91.75 107.5 65

1.90 30.60 184 27.90 0.00 11.3 0.00 4.99 3942 4673 252 0.00 955 90 19.33 27.97 7066 23318 184 910 0.00 16.75 23.71 1168 16260 137 0.00 50 1.09 5.46 0.00 80.93 224 0.00 0.00 4.91 4014 55.76 1302 0.00 525 6.5 2.15 6.10 3.54 1.85 0.00 0.00 0.00 14.60 889 1400 13637 76.09 115 9 3.15 5477 5319 9383 0.00 660 80 5.42 13555 8813 2386 82.47 1490 250 4.29 28.68 23.05 25.43 0.00 140 36.5 Continued

Table B3 continued

15.68 245 4719 9460 0.00 86 0 0.70 3.69 1.22 0.00 82.61 0 0.00 9.59 9311 4779 8320 0.00 880 300 23.46 13.94 7921 42614 314 180 700 1.30 4.53 17.82 0.00 100 1 0.00 9.32 10272 6381 6438 0.00 1240 185 12.84 1196 6.23 521 89.02 4.27 250 3.06 9.38 1.41 0.86 510 0.00 0.00 3.46 9.47 3.42 2.12 620 0.00 0.00 33.57 9.79 1721 28321 173 915 550

216 1.72 5.99 0.96 0.68 0.00 0.00 0.00

19.42 17.17 1502 14328 0.00 1950 240 43.59 3170 38435 15886 112 312.5 20 3.11 12.89 3.85 6.36 218 0.00 0.00 2.18 6.13 1.05 0.26 149 0.00 0.00 120 14.01 12967 310 393 75000 6750 5.41 343 1053 780 0.00 79.5 0 0.00 2.56 0.00 3.16 266 0.00 0.00 0.00 7.39 0.00 8.47 414 0.00 0.00 1.74 5.89 1.16 0.97 0.00 0.00 0.00 0.00 200 80.31 39.06 0.00 34.55 0.00 1.08 12.49 8.61 0.00 0.00 3.5 1 Continued

Table B3 continued

3.66 4.70 0.00 0.00 701 0.00 0.00 1.89 7.08 1.62 0.36 135 0.5 0.00 62.74 9.67 1942 55266 288 5400 3100 0.00 4.44 0.00 10.43 2644 0.00 0.00 3.45 135 3.52 107 0.00 2.79 0.00 4.16 68.96 37.30 12.72 190 0.00 0.00 3.81 168 70.36 990 0.00 24 5.5 1.33 5.46 6.22 0.00 138 0.00 0.00 3.57 198 106 63.86 0.00 23 0.00 3.10 7.81 1.35 0.00 0.00 0.00 0.00

217 0.00 3.64 0.46 0.00 0.00 0 0.00

6.55 429 418 119 0.00 137 0.00 15.84 4827 4841 7670 109 21.1 0.00 4.04 7.60 10.40 65.24 109 6.2 65 0.00 3.10 3.49 5.47 0.00 0.00 0.00 Continued

Table B3 continued

2+ -1 2+ -1 + -1 + -1 2+ -1 -1 Ca (µg g ) Mg (µg g ) Na (µg g ) K (µg g ) Sr (µg g ) NH3 (µg kg ) N:P ratio Total salt (µg (molar) g-1) 4.39 0.14 0.39 5.06 0.00 123 0.00 16.36 4.37 0.16 0.39 5.06 0.00 39.75 0.00 18.38 2.06 0.91 1.17 1.78 0.00 173 0.00 12.84 3.37 1.54 2.12 3.01 0.00 167 0.00 13.47 3.28 1.39 1.74 2.19 0.00 37.81 7.68 17.12 5.55 2.50 2.50 8.90 0.00 683 114 40.60 4.48 1.58 1.14 2.66 0.00 80.10 33.86 19.33 6.24 1.30 55.92 14.19 0.00 142 0.00 172 39.14 0.64 2.44 2.99 0.00 238 0.00 50.52

218 3.83 1.74 17.00 5.70 0.00 404 4.91 36.49

76.57 1.44 5.14 7.86 0.00 382 8.98 108 90.40 12.30 49.00 10.75 0.00 70.48 0.00 228 4.31 1.53 1.89 2.40 0.00 98.94 0.00 22.37 7.88 0.40 1.45 1.67 0.00 222 0.00 20.38 4.61 1.39 9.26 2.76 0.00 244 0.00 29.11 64.25 13.55 175 8.95 0.00 35.69 0.00 860 0.55 0.00 16.44 0.00 0.00 33.48 68.98 43.89 40.75 20.95 56.94 5.45 0.00 37.46 0.00 351 12.17 7.76 95.83 3.91 0.07 203 9217 804 15.85 4.65 15.75 12.60 0.00 103 1.73 67.95 Continued

Table B3 continued

17.40 5.80 166 7.60 0.00 128 0.00 614 428 47.20 614 24.85 1.20 99.89 0.00 3664 61.75 37.40 77.53 5.45 0.35 277 0.00 547 0.55 0.13 16.19 0.00 0.00 59.45 0.00 28.13 1.44 0.38 53.08 1.56 0.00 44.21 0.00 87.98 46.37 21.61 760 0.00 0.00 76.02 0.00 2282 61.50 27.10 48.96 5.65 0.45 184 0.00 389 94.10 14.95 39.03 4.63 0.00 50.51 0.00 205 37.96 6.92 316 7.68 0.00 49.38 0.00 536 312 234 294 14.30 1.25 38.33 0.00 3541

219 0.77 0.68 25.56 0.95 0.00 74.09 0.00 56.12

759 92.24 449 0.00 0.00 37.19 16926 5242 6.80 3.05 4.33 6.40 0.00 134 91.72 36.57 351 83.55 401 5.20 0.85 34.29 0.00 2874 46.37 13.82 391 0.00 0.00 92.59 0.00 1796 484 35.95 1989 5.60 0.40 56.74 24103 9239 381 57.65 525 13.90 0.65 61.30 0.00 1370 538 59.11 308 26.74 0.86 139 0.00 1197 3745 546 4985 79.99 9.99 738 33.79 37797 171 70.32 196 9.33 0.76 56.46 0.00 798 63.34 14.11 214 10.20 0.00 140 0.00 447 630 59.74 591 14.57 0.96 55.38 0.00 1756 Continued

Table B3 continued

9.50 6.20 49.40 3.50 0.00 74.88 0.00 198 5.41 1.58 90.96 6.82 0.00 92.40 80.72 197 1284 76.73 426 20.75 1.90 110 0.00 2251 148 83.10 286 7.06 0.39 47.35 0.00 1879 6.20 1.40 11.20 4.61 0.00 87.07 3.70 35.66 139 25.85 222 8.65 0.50 184 0.00 1327 0.67 0.42 62.92 0.86 0.00 508 49.86 161 463 321 597 20.70 1.25 47.81 0.00 5984 298 229 735 23.85 2.15 345 0.00 4484 514 511 1913 35.40 2.50 420 0.00 11672

220 24.05 7.15 93.60 5.40 0.00 31.78 0.00 378

357 79.31 122 8.89 0.72 43.18 0.00 779 1492 102 4453 59.32 5.16 268 5756 20914 6.60 0.85 9.35 21.23 0.00 344 17767 41573 657 72.58 4131 38.15 3.74 32.59 18.34 16720 1357 103 1935 37.56 2.35 200 0.00 12392 339 183 1144 18.79 2.15 95.27 0.00 6477 1078 303 349 23.15 2.50 87.77 0.00 6788 1191 175 2096 59.95 4.85 189 0.00 17905 111 110 403 14.85 0.95 32.93 0.00 2051 3.89 1.25 270 2.33 0.00 169 39.72 520 684 486 737 18.42 1.70 210 0.00 5516 Continued

Table B3 continued

104 117 377 15.05 1.00 104 0.00 3040 5.38 3.14 78.75 2.83 0.00 155 0.00 250 155 121 287 14.05 1.05 73.46 0.00 1336 1929 339 869 23.51 3.61 67.22 0.00 11050 1317 124 81.75 7.64 0.00 60.22 0.00 5725 2.96 0.97 1.39 2.04 0.00 55.20 0.00 13.12 1592 366 395 21.85 13.60 148 0.00 4989 67.42 20.11 97.77 5.47 0.00 33.31 0.00 524 2.90 2.95 31.05 3.65 0.00 18.85 0.00 105 659 257 790 26.25 2.05 110 0.00 3848

221 748 660 833 23.74 2.37 279 0.00 7529

3.80 2.45 8.65 5.25 0.00 111 0.00 26.33 2094 275 435 20.75 5.90 212 0.00 11276 14.00 12.90 115 6.85 0.00 690 1184 397 1487 213 552 12.10 1.35 114 0.00 2292 2745 579 11134 103 11.09 197 3926 43245 15.20 4.45 237 3.80 0.00 277 113 593 36.10 7.85 338 10.70 0.00 163 68.79 981 142 39.45 193 7.50 0.45 49.82 0.00 1458 1947 424 11886 122 9.25 722 18393 51090 3.85 0.12 240 3.55 0.00 85.25 39.36 367 1694 396 363 24.30 2.40 213 0.00 4667 Continued

Table B3 continued

6.20 0.53 1.71 1.26 0.00 166 0.00 17.19 2252 157 715 30.98 5.04 165 1466 11665 2686 343 5182 59.00 8.70 111 5526 23370 4.90 2.85 10.45 6.75 0.00 221 1.35 36.03 2506 446 977 43.90 19.85 1142 19150 16856 1.30 0.90 93.03 1.91 0.00 75.30 0.00 254 78.75 18.20 169 7.55 0.00 84.91 0.00 350 4.89 3.04 16.30 2.32 0.00 20.35 0.00 113 1234 506 498 36.30 6.00 62.37 0.00 8088 2068 750 839 35.29 3.64 901 0.00 9060

222 199 41.01 431 5.86 0.36 33.31 0.00 2184

0.68 0.41 9.18 1.02 0.01 55.96 0.00 22.97 79.52 28.04 103 8.25 0.00 220 87.32 705 2505 133 555 40.60 3.75 43.83 16.84 11464 2971 394 1343 15.65 9.75 176 0.00 18116 15.76 5.74 89.22 4.34 0.00 49.49 13344 305 2089 489 474 41.50 3.90 49.44 0.00 9278 1590 1277 2906 85.81 12.96 408 88038 27714 6.58 1.84 8.29 3.25 0.00 102 12.31 29.53 387 262 707 20.30 2.50 231 0.00 5003 3180 104 255 21.60 3.35 523 0.00 13816 1409 1083 3226 54.13 10.50 101 0.00 19863 Continued

Table B3 continued

92.23 66.94 197 7.64 0.62 42.86 0.00 1315 3001 817 5883 37.83 8.54 168 76894 33981 1458 424 709 80.70 5.30 543 0.00 11523 527 747 1366 60.80 6.70 184 106043 10608 4404 698 1419 32.64 6.91 341 17759 16427 1164 440 445 33.00 4.45 97.30 0.00 4820 4053 897 1797 24.52 10.20 105 14005 18952 6.90 1.36 3.08 2.93 0.00 167 0.00 23.75 2685 953 1955 124 32.90 1145 81574 23198 354 18.20 113 18.55 0.75 40.74 0.00 1544

223 3.65 1.73 7.70 4.11 0.00 74.34 84.91 43.83

2131 480 1005 29.79 6.11 93.40 0.00 11537 1869 698 874 46.10 11.50 3190 0.00 14063 6.90 1.36 3.08 2.93 0.00 263 45.28 27.78 50.10 8.20 10.20 25.40 0.00 561 0.00 335 2244 293 1007 80.05 10.65 960 27539 13738 0.78 0.58 14.57 1.07 0.01 107 0.00 31.30 1992 698 1166 78.25 16.35 2702 59419 16289 1203 280 354 11.70 17.40 84.12 0.00 6963 1717 577 1076 69.30 23.55 5077 0.00 10325 1702 419 875 21.84 4.20 98.71 67906 10998 3234 1053 6086 36.93 16.90 164 80591 33465 Continued

Table B3 continued

756 254 1026 50.80 3.70 2606 0.00 7932 35.75 6.45 168 6.40 0.00 177 160 444 2827 840 2374 68.40 16.30 166 0.00 23789 5.30 2.35 1.60 4.50 0.00 150 0.00 23.31 2016 2514 2642 65.74 22.24 1242 19096 30520 2280 823 2663 133 34.34 1702 28669 22702 4109 150 869 18.21 3.69 106 19932 18529 3098 569 1222 45.74 30.52 288 0.00 19979 11.20 0.66 11.13 2.64 0.00 109 0.00 40.19 21.85 4.65 16.85 10.85 0.00 65.12 83.61 74.67

224 2803 158 1146 94.49 27.35 2336 666 15638

60.80 19.30 77.38 7.22 0.00 31.35 0.00 410 1662 405 2251 41.55 9.80 252 0.00 13243 2309 1377 4712 148 32.87 545 58864 39011 2553 842 2557 139 34.55 1055 13103 23594 29.00 3.45 17.05 5.10 0.00 54.22 0.00 142 1360 469 1515 36.70 7.30 387 0.00 8765 18.50 8.20 6.45 7.90 0.00 1078 0.00 54.69 2865 215 521 57.40 5.62 139 28191 19605 3394 752 2615 45.25 9.49 158 0.00 26998 2598 1226 6340 74.60 25.36 170 163698 35026 15.30 4.15 39.30 4.50 0.00 454 61028 145 Continued

Table B3 continued

3512 645 1570 20.09 0.00 69.51 0.00 20187 5.40 2.30 1.90 5.75 0.00 166 22.69 21.05 4316 910 3678 79.55 20.25 185 0.00 31424 2828 5781 16958 268 46.61 1012 38620 76455 3.94 2.03 5.37 2.65 0.00 73.28 272 37.75 3726 1067 3979 59.20 21.49 256 0.00 31953 26.45 5.40 51.95 5.05 0.00 192 107 1825 3.23 1.74 9.97 3.96 0.00 78.70 4.25 34.12 5.15 3.05 13.70 6.50 0.00 67.88 8.46 47.50 1934 1800 7470 171 32.85 715 15230 41494

225 4.40 2.15 2.20 3.95 0.00 121 0.00 22.06

2277 240 3063 83.77 24.62 410 0.00 21556 3384 1412 4032 130 38.65 383 525564 66532 48.86 2.08 38.26 4.90 0.08 73.15 27.03 121 7.30 3.20 4.43 6.65 0.00 467 10.78 31.34 1913 6281 25260 440 36.49 2845 50567 47417 2630 275 728 13.70 8.80 185 0.00 5838 39.52 0.76 4.09 2.13 0.00 287 0.00 52.49 6.82 1.23 22.85 4.52 0.00 52.36 0.00 51.70 2.91 1.70 3.53 2.80 0.00 58.96 0.00 20.70 12.00 4.80 178 4.55 0.00 97.77 0.00 519 17.72 2.37 7.20 5.06 0.00 1229 0.00 54.54 Continued

Table B3 continued

3.45 2.15 15.45 5.60 0.00 89.83 0.00 35.71 31.88 0.76 0.92 2.98 0.00 293 18.35 47.62 1717 1750 19301 395 31.89 538 10336 80482 87.60 2.45 6.10 9.10 0.00 1908 0.00 123 15.35 4.41 39.22 1.41 0.00 149 0.00 310 6.35 4.80 71.50 4.50 0.00 193 301 210 322 41.75 157 12.05 1.05 399 0.00 1766 6.90 3.25 11.20 6.60 0.00 72.91 68.94 41.10 29.50 8.75 181 5.45 0.00 78.43 0.00 596 6.10 3.50 8.25 6.15 0.00 220 0.00 36.26

226 3.32 1.08 1.83 4.86 0.00 642 0.00 15.19

62.50 69.10 282 12.35 0.60 548 0.00 1400 1954 853 698 41.50 3.11 147 39.72 20904 35.85 7.00 60.50 4.65 0.00 104 146 195 4.04 0.87 7.34 3.48 0.00 118 0.00 27.79

Table B4. Random forest model imputed testing dataset.

Sample ID Latitude Longitude Elevation (m) Distance from Distance from Soil moisture coast (km) glacier (m) (%) MSP1-2 -84.4831 -176.5035 264 1 408 0.16 MF2-4 -84.6237 -176.7252 424 9 218 0.39 MF1-5 -84.6265 -176.7498 531 9 350 0.53 MSP4-2B -84.4657 -177.1357 287 1 11 0.60 RM2-5 -85.4868 -177.1639 1754 120 706 0.63 RM1-8 -85.4868 -177.0994 1731 120 653 0.69 AV1-2 -85.1762 -174.0922 1492 72 1097 0.72 TGV1-7 -84.9161 -176.8989 982 45 267 0.76 RM1-1 -85.4909 -177.1772 1801 120 1435 0.84

227 BP2-7 -85.2036 -177.3920 1235 82 93 0.89

TGV3-8 -84.9136 -176.8835 912 45 74 0.96 BP2-5A -85.2072 -177.3887 1294 82 396 0.98 MSP4-2C -84.4657 -177.1357 287 1 11 1.05 AV2-4 -85.1696 -174.1364 1369 72 275 1.11 MSP3-1 -84.4819 -176.4879 150 1 214 1.23 MW 4-2 -84.5596 -176.8112 352 10 1 1.31 BP3-1 -85.2144 -177.4048 1422 82 1041 1.43 MH1-7 -85.0310 -177.1790 1058 63 204 1.52 AV2-2 -85.1703 -174.1350 1405 72 352 1.60 TN1-5 -84.9210 -176.0976 994 45 66 1.74 Continued

Table B4 continued

MH1-3 -85.0419 -177.2421 1196 63 1540 1.93 TGV1-2 -84.9199 -176.9831 1198 45 1186 2.07 MF3-2 -84.6227 -176.7456 520 9 454 2.33 TN1-1 -84.9257 -176.0692 1137 45 168 2.58 TN2-7 -84.9266 -176.1094 1068 45 343 2.90 SH3-2 -85.3597 -175.0693 2137 94 900 3.01 TN2-6 -84.9266 -176.1063 1068 45 320 3.57 NP3-1 -84.5346 -175.4232 683 1 66 3.99 MFX-2 -84.6241 -176.7085 329 9 34 4.85 TN2-2 -84.9242 -176.0996 1034 45 76 6.43

228 AV3-8 -85.1674 -174.1452 1351 72 35 9.38

Continued

Table B4 continued

- -1 - -1 - -1 2- -1 3- -1 - -1 - -1 F (µg g ) Cl (µg g ) NO3 (µg g ) SO4 (µg g ) PO4 (µg kg ) ClO4 (µg kg ) ClO3 (µg kg ) 0.00 6.03 0.31 0.00 0.00 0.00 0.00 1.00 10.18 8.06 0.00 0.00 0.35 0.00 0.00 3.65 0.32 4.78 245 0.00 0.00 0.00 27.14 0.79 0.00 4195 0.00 0.00 10.86 3.62 351 3419 0.00 205 1650 7.25 1.59 12.66 290 0.00 44.55 1250 2.14 124 666 13.27 0.00 130 18 5.96 21.46 19.98 37.95 324 8.2 0.00 9.51 4.84 49.12 7902 0.00 430 2200 229 5.02 4.13 50.19 13.25 825 0.00 0.00

4.09 22.12 25.22 0.00 0.00 7.85 0.00 5.35 790 707 4087 0.00 41.5 48.5 3.37 7.21 3.60 1.79 1052 0.00 0.00 6.40 184 956 7603 0.00 110.5 4.5 2.84 4.70 0.10 0.00 0.00 3.915 0.00 0.89 4.31 0.54 0.00 0.00 0.00 0.00 16.18 960 2374 9190 231 103 80 8.33 435 471 201 100 102.5 0.00 4.36 215 1593 474 0.00 150.5 40 3.85 2125 1376 226 0.00 263 0.00 Continued

Table B4 continued

13.95 1904 1691 8594 95.73 470 0 16.10 3888 6602 263 0.00 3975 0 2.72 6.43 1.84 0.72 524 0.00 0.00 8.30 5932 6680 4988 100 1410 270 6.68 5934 8363 941 0.00 680 0 9.15 8.00 7551 21009 190 3550 6750 2.05 829 521 114 0.00 113 5 0.00 3.02 1.54 9.13 0.00 0.00 0.00 3.27 6.88 1.37 0.00 522 0.00 0.00 0.84 11.70 2.34 65.24 0.00 0.00 0.00

230 4.79 15.32 23.32 84.67 0.00 8.6 0.00

Continued

Table B4 continued

2+ -1 2+ -1 + -1 + -1 2+ -1 -1 Ca (µg g ) Mg (µg g ) Na (µg g ) K (µg g ) Sr (µg g ) NH3 (µg kg ) N:P ratio Total salt (µg (molar) g-1) 1.97 0.88 1.50 1.89 0.00 54.13 0.00 12.59 1.70 1.08 1.48 2.04 0.00 131 0.00 25.54 2.80 2.00 5.65 4.70 0.00 135 2.00 24.14 6.50 1.44 4.25 4.30 0.00 88.41 0.29 48.61 552 80.00 818 20.30 1.40 49.80 0.00 5256 8.85 2.20 132 4.90 0.00 134 0.00 460 50.17 85.68 219 10.31 0.38 54.63 0.00 1172 2.46 2.92 48.82 2.22 0.00 95.04 94.35 142 1098 53.60 1896 34.50 2.50 200 0.00 11050

231 3.85 0.12 24.39 0.00 0.00 99.71 93.24 102

2.66 2.06 37.01 2.16 0.00 141 0.00 95.32 2770 555 1174 18.31 0.00 109 0.00 10107 99.36 1.96 10.92 7.20 0.09 152 5.25 137 2211 364 398 20.09 2.03 78.39 0.00 11745 5.00 2.70 4.40 8.70 0.00 976 0.00 28.45 7.20 3.05 6.85 3.55 0.00 80.20 0.00 26.39 2332 421 1610 30.80 4.60 123 15762 16938 54.15 34.64 431 8.86 0.00 37.89 7195 1645 1738 382 323 24.34 5.07 178 0.00 4759 638 166 817 18.20 6.40 34.81 0.00 5377 Continued

Table B4 continued

2801 902 1074 43.35 8.45 824 27054 17032 1527 1605 3791 71.73 15.85 109 0.00 17784 4.25 1.90 9.10 5.15 0.00 44.53 5.38 32.63 3801 808 3665 46.52 15.83 383 101855 25948 4157 539 4605 36.17 4.96 258 0.00 24588 2674 3129 3836 56.11 14.40 1035 60788 38290 265 96.80 318 14.50 1.05 77.76 0.00 2161 6.33 0.94 5.57 2.24 0.00 167 0.00 28.77 7.15 3.95 12.30 6.55 0.00 291 4.03 41.99 9.81 1.13 15.87 1.50 0.00 72.66 0.00 108

232 8.60 5.30 60.60 3.95 0.00 265 0.00 207

Figure B1. Images of biology at Shackleton Glacier terminus

Images of black lichen (a), green lichen (b), collembola (c), and multiple lichen species (d) at Mt. Speed and Mt. Wasko.

233

Code B1. R code for multiple linear regression and random forest models

#Multiple linear regression and random forest models for prediction of salt concentrations in Antarctic soils #The training datasets are published by Diaz et al. (2020): #"Geochemical zones and environmental gradients for soils from the Central Transantarctic Mountains #Please cite Diaz et al. (2020) for use of the datasets and models

#Loading all required packages library(ggplot2) library(randomForest) library(e1071) library(caret) library(smatr)

#Import the training datasets for the multiple linear regression and random forest models Salts <- read_excel("~/Downloads/Salts.xlsx") RFSalts_Train <- read_excel("~/Downloads/RFSalts_Train.xlsx") RFSalts_Test <-read_excel("~/Downloads/RFSalts_Test.xlsx") Darwin <- read_excel("~/Downloads/Darwin_Test.xlsx")

######################################################################## ################################################### #Creating the multiple linear regression models using the Salts dataset

#For Total salts #Creating the linear model with the 5 geographic parameters + soil moisture TSreg <- lm(data = RFSalts_Train, Total_salt~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture) #Printing model statistics to evaluate model strength summary(TSreg)

#For N:P ratio NPreg <- lm(data = RFSalts_Train, NP~ Latitude + Longitude + Elevation + Distance_from_coast + 234

Distance_from_glacier + Soil_moisture) summary(NPreg)

#For ClO4 ClO4reg <- lm(data = Salts, ClO4~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture) summary(ClO4reg)

#For Cl Clreg <- lm(data = Salts, Cl~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture) summary(Clreg)

#For F Freg <- lm(data = Salts, F~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture) summary(Freg)

#For NO3 NO3reg <- lm(data = Salts, NO3~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture) summary(NO3reg)

#For SO4 SO4reg <- lm(data = Salts, SO4~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture) summary(SO4reg)

#For PO4 PO4reg <- lm(data = Salts, PO4~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture) summary(PO4reg)

#For Ca

235

Careg <- lm(data = Salts, Ca~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture) summary(Careg)

#For Mg Mgreg <- lm(data = Salts, Mg~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture) summary(Mgreg)

#For Na Nareg <- lm(data = Salts, Na~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture) summary(Nareg)

#For K Kreg <- lm(data = Salts, K~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture) summary(Kreg)

#For Sr Srreg <- lm(data = Salts, Sr~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture) summary(Srreg)

#For NH3 NH3reg <- lm(data = Salts, NH3~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture) summary(NH3reg)

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

#Creating the random forest model with the training dataset

236

#mtry is set as default, which in this case is n/3 = 2. This is the value with the minimum error

#For Total salts #Building the random forest model RFModelTS <- randomForest(Total_salt ~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture, data = RFSalts_Train, na.action = na.omit, ntree = 2000) #Seeing which variables were most and least important for nude purity varImpPlot(RFModelTS)

#For N:P molar ratio RFModelNP <- randomForest(NP ~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture, data = RFSalts_Train, na.action = na.omit, ntree = 2000) varImpPlot(RFModelNP)

#For ClO4 RFModelClO4 <- randomForest(ClO4 ~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture, data = RFSalts_Train, na.action = na.omit, ntree = 2000) varImpPlot(RFModelClO4)

#For Cl RFModelCl <- randomForest(Cl ~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture, data = RFSalts_Train, na.action = na.omit, ntree = 2000) varImpPlot(RFModelCl)

#For F RFModelF <- randomForest(F ~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture, data = RFSalts_Train, na.action = na.omit, ntree = 2000) varImpPlot(RFModelF)

#For NO3

237

RFModelNO3 <- randomForest(NO3 ~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture, data = RFSalts_Train, na.action = na.omit, ntree = 2000) varImpPlot(RFModelNO3)

#For SO4 RFModelSO4 <- randomForest(SO4 ~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture, data = RFSalts_Train, na.action = na.omit, ntree = 2000) varImpPlot(RFModelSO4)

#For PO4 RFModelPO4 <- randomForest(PO4 ~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture, data = RFSalts_Train, na.action = na.omit, ntree = 2000) varImpPlot(RFModelPO4)

#For Ca RFModelCa <- randomForest(Ca ~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture, data = RFSalts_Train, na.action = na.omit, ntree = 2000) varImpPlot(RFModelCa)

#For Mg RFModelMg <- randomForest(Mg ~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture, data = RFSalts_Train, na.action = na.omit, ntree = 2000) varImpPlot(RFModelMg)

#For Na RFModelNa <- randomForest(Na ~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture, data = RFSalts_Train, na.action = na.omit, ntree = 2000) varImpPlot(RFModelNa)

238

#For K RFModelK <- randomForest(K ~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture, data = RFSalts_Train, na.action = na.omit, ntree = 2000) varImpPlot(RFModelK)

#For Sr RFModelSr <- randomForest(Sr ~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture, data = RFSalts_Train, na.action = na.omit, ntree = 2000) varImpPlot(RFModelSr)

#For NH3 RFModelNH3 <- randomForest(NH3 ~ Latitude + Longitude + Elevation + Distance_from_coast + Distance_from_glacier + Soil_moisture, data = RFSalts_Train, na.action = na.omit, ntree = 2000) varImpPlot(RFModelNH3) ######################################################################## ###################################################

#Running the Shackleton Glacier region test dataset on the generated multiple linear regression models

#For Total salts #Inputting the new test dataset and storing the predicted concentration results TSLMPred <- predict(TSreg, newdata = RFSalts_Test) #Running a correlation between the measured and predicted values cor.test(TSLMPred, RFSalts_Test$Total_salt) #Quick visual check of the data plot(TSLMPred, RFSalts_Test$Total_salt) #Determining the slope by fitting bivariate lines using the standardized major axis (SMA) for measured and predicted values #A slope of 1 is ideal, where there is a direct 1:1 relationship #sma will test the slope estimate and determine the test statistic for the null hypothesis sma(TSLMPred~ RFSalts_Test$Total_salt, slope.test=1) #Determing the mean absolute error (MAE) and root mean squared error (RMSE) mae(TSLMPred, RFSalts_Test$Total_salt)

239

rmse(TSLMPred, RFSalts_Test$Total_salt)

#For N:P ratio NPLMPred <- predict(NPreg, newdata = RFSalts_Test) cor.test(NPLMPred, RFSalts_Test$NP) plot(NPLMPred, RFSalts_Test$NP) sma(NPLMPred~ RFSalts_Test$NP, slope.test=1) mae(NPLMPred, RFSalts_Test$NP) rmse(NPLMPred, RFSalts_Test$NP)

#For ClO4 ClO4LMPred <- predict(ClO4reg, newdata = RFSalts_Test) cor.test(ClO4LMPred, RFSalts_Test$ClO4) plot(ClO4LMPred, RFSalts_Test$ClO4) sma(ClO4LMPred~ RFSalts_Test$ClO4, slope.test=1) mae(ClO4LMPred, RFSalts_Test$ClO4) rmse(ClO4LMPred, RFSalts_Test$ClO4)

#For ClO3 ClO3LMPred <- predict(ClO3reg, newdata = RFSalts_Test) cor.test(ClO3LMPred, RFSalts_Test$ClO3) plot(ClO3LMPred, RFSalts_Test$ClO3) sma(ClO3LMPred~ RFSalts_Test$ClO3, slope.test=1) mae(ClO3LMPred, RFSalts_Test$ClO3) rmse(ClO3LMPred, RFSalts_Test$ClO3)

#For Cl ClLMPred <- predict(Clreg, newdata = RFSalts_Test) cor.test(ClLMPred, RFSalts_Test$Cl) plot(ClLMPred, RFSalts_Test$Cl) sma(ClLMPred~ RFSalts_Test$Cl, slope.test=1) mae(ClLMPred, RFSalts_Test$Cl) rmse(ClLMPred, RFSalts_Test$Cl)

#For F FLMPred <- predict(Freg, newdata = RFSalts_Test) cor.test(FLMPred, RFSalts_Test$F) plot(FLMPred, RFSalts_Test$F)

240

sma(FLMPred~ RFSalts_Test$F, slope.test=1) mae(FLMPred, RFSalts_Test$F) rmse(FLMPred, RFSalts_Test$F)

#For NO3 NO3LMPred <- predict(NO3reg, newdata = RFSalts_Test) cor.test(NO3LMPred, RFSalts_Test$NO3) plot(NO3LMPred, RFSalts_Test$NO3) sma(NO3LMPred~ RFSalts_Test$NO3, slope.test=1) mae(NO3LMPred, RFSalts_Test$NO3) rmse(NO3LMPred, RFSalts_Test$NO3)

#For SO4 SO4LMPred <- predict(SO4reg, newdata = RFSalts_Test) cor.test(SO4LMPred, RFSalts_Test$SO4) plot(SO4LMPred, RFSalts_Test$SO4) sma(SO4LMPred~ RFSalts_Test$SO4, slope.test=1) mae(SO4LMPred, RFSalts_Test$SO4) rmse(SO4LMPred, RFSalts_Test$SO4)

#For PO4 PO4LMPred <- predict(PO4reg, newdata = RFSalts_Test) cor.test(PO4LMPred, RFSalts_Test$PO4) plot(PO4LMPred, RFSalts_Test$PO4) sma(PO4LMPred~ RFSalts_Test$PO4, slope.test=1) mae(PO4LMPred, RFSalts_Test$PO4) rmse(PO4LMPred, RFSalts_Test$PO4)

#For Ca CaLMPred <- predict(Careg, newdata = RFSalts_Test) cor.test(CaLMPred, RFSalts_Test$Ca) sma(CaLMPred~ RFSalts_Test$Ca, slope.test=1) mae(CaLMPred, RFSalts_Test$Ca) rmse(CaLMPred, RFSalts_Test$Ca)

#For Mg MgLMPred <- predict(Mgreg, newdata = RFSalts_Test) cor.test(MgLMPred, RFSalts_Test$Mg)

241

plot(MgLMPred, RFSalts_Test$Mg) sma(MgLMPred~ RFSalts_Test$Mg, slope.test=1) mae(MgLMPred, RFSalts_Test$Mg) rmse(MgLMPred, RFSalts_Test$Mg)

#For Na NaLMPred <- predict(Nareg, newdata = RFSalts_Test) cor.test(NaLMPred, RFSalts_Test$Na) plot(NaLMPred, RFSalts_Test$Na) sma(NaLMPred~ RFSalts_Test$Na, slope.test=1) mae(NaLMPred, RFSalts_Test$Na) rmse(NaLMPred, RFSalts_Test$Na)

#For K KLMPred <- predict(Kreg, newdata = RFSalts_Test) cor.test(KLMPred, RFSalts_Test$K) plot(KLMPred, RFSalts_Test$K) sma(KLMPred~ RFSalts_Test$K, slope.test=1) mae(KLMPred, RFSalts_Test$K) rmse(KLMPred, RFSalts_Test$K)

#For Sr SrLMPred <- predict(Srreg, newdata = RFSalts_Test) cor.test(SrLMPred, RFSalts_Test$Sr) plot(SrLMPred, RFSalts_Test$Sr) sma(SrLMPred~ RFSalts_Test$Sr, slope.test=1) mae(SrLMPred, RFSalts_Test$Sr) rmse(SrLMPred, RFSalts_Test$Sr)

#For NH3 NH3LMPred <- predict(NH3reg, newdata = RFSalts_Test) cor.test(NH3LMPred, RFSalts_Test$NH3) plot(NH3LMPred, RFSalts_Test$NH3) sma(NH3LMPred~ RFSalts_Test$NH3, slope.test=1) mae(NH3LMPred, RFSalts_Test$NH3) rmse(NH3LMPred, RFSalts_Test$NH3)

242

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

#Running the Shackleton Glacier region test dataset on the generated random forest models

#For Total salts #Inputting the new test dataset and storing the predicted concentration results TSRFPred<- predict(RFModelTS, newdata = RFSalts_Test) #Running a correlation between the measured and predicted values cor.test(TSRFPred, RFSalts_Test$Total_salt) #Quick visual check of the data plot(TSRFPred, RFSalts_Test$Total_salt) #Determining the slope by fitting bivariate lines using the standardized major axis (SMA) for measured and predicted values #A slope of 1 is ideal, where there is a direct 1:1 relationship #sma will test the slope estimate and determine the test statistic for the null hypothesis sma(TSRFPred ~ RFSalts_Test$Total_salt, slope.test=1) #Determing the mean absolute error (MAE) and root mean squared error (RMSE) mae(TSRFPred, RFSalts_Test$Total_salt) rmse(TSRFPred, RFSalts_Test$Total_salt)

#For N:P ratio NPRFPred<- predict(RFModelNP, newdata = RFSalts_Test) cor.test(NPRFPred, RFSalts_Test$NP) plot(NPRFPred, RFSalts_Test$NP) sma(NPRFPred ~ RFSalts_Test$NP, slope.test=1) mae(NPRFPred, RFSalts_Test$NP) rmse(NPRFPred, RFSalts_Test$NP)

#For ClO4 ClO4RFPred<- predict(RFModelClO4, newdata = RFSalts_Test) cor.test(ClO4RFPred, RFSalts_Test$ClO4) plot(ClO4RFPred, RFSalts_Test$ClO4) sma(ClO4RFPred ~ RFSalts_Test$ClO4, slope.test=1) mae(ClO4RFPred, RFSalts_Test$ClO4) rmse(ClO4RFPred, RFSalts_Test$ClO4)

243

#For ClO3 ClO3RFPred<- predict(RFModelClO3, newdata = RFSalts_Test) cor.test(ClO3RFPred, RFSalts_Test$ClO3) plot(ClO3RFPred, RFSalts_Test$ClO3) sma(ClO3RFPred ~ RFSalts_Test$ClO3, slope.test=1) mae(ClO3RFPred, RFSalts_Test$ClO3) rmse(ClO3RFPred, RFSalts_Test$ClO3)

#For Cl ClRFPred<- predict(RFModelCl, newdata = RFSalts_Test) cor.test(ClRFPred, RFSalts_Test$Cl) plot(ClRFPred, RFSalts_Test$Cl) sma(ClRFPred ~ RFSalts_Test$Cl, slope.test=1) mae(ClRFPred, RFSalts_Test$Cl) rmse(ClRFPred, RFSalts_Test$Cl)

#For K KRFPred<- predict(RFModelK, newdata = RFSalts_Test) cor.test(KRFPred, RFSalts_Test$K) plot(KRFPred, RFSalts_Test$K) sma(KRFPred ~ RFSalts_Test$K, slope.test=1) mae(KRFPred, RFSalts_Test$K) rmse(KRFPred, RFSalts_Test$K)

#For NO3 NO3RFPred<- predict(RFModelNO3, newdata = RFSalts_Test) cor.test(NO3RFPred, RFSalts_Test$NO3) sma(NO3RFPred ~ RFSalts_Test$NO3, slope.test=1) mae(NO3RFPred, RFSalts_Test$NO3) rmse(NO3RFPred, RFSalts_Test$NO3)

#For SO4 SO4RFPred<- predict(RFModelSO4, newdata = RFSalts_Test) cor.test(SO4RFPred, RFSalts_Test$SO4) plot(SO4RFPred, RFSalts_Test$SO4) sma(SO4RFPred ~ RFSalts_Test$SO4, slope.test=1) mae(SO4RFPred, RFSalts_Test$SO4) rmse(SO4RFPred, RFSalts_Test$SO4)

244

#For PO4 PO4RFPred<- predict(RFModelPO4, newdata = RFSalts_Test) cor.test(PO4RFPred, RFSalts_Test$PO4) plot(PO4RFPred, RFSalts_Test$PO4) sma(PO4RFPred ~ RFSalts_Test$PO4, slope.test=1) mae(PO4RFPred, RFSalts_Test$PO4) rmse(PO4RFPred, RFSalts_Test$PO4)

#For Ca CaRFPred<- predict(RFModelCa, newdata = RFSalts_Test) cor.test(CaRFPred, RFSalts_Test$Ca) plot(CaRFPred, RFSalts_Test$Ca) sma(CaRFPred ~ RFSalts_Test$Ca, slope.test=1) mae(CaRFPred, RFSalts_Test$Ca) rmse(CaRFPred, RFSalts_Test$Ca)

#For Mg MgRFPred<- predict(RFModelMg, newdata = RFSalts_Test) cor.test(MgRFPred, RFSalts_Test$Mg) plot(MgRFPred, RFSalts_Test$Mg) sma(MgRFPred ~ RFSalts_Test$Mg, slope.test=1) mae(MgRFPred, RFSalts_Test$Mg) rmse(MgRFPred, RFSalts_Test$Mg)

#For Na NaRFPred<- predict(RFModelNa, newdata = RFSalts_Test) cor.test(NaRFPred, RFSalts_Test$Na) plot(NaRFPred, RFSalts_Test$Na) sma(NaRFPred ~ RFSalts_Test$Na, slope.test=1) mae(NaRFPred, RFSalts_Test$Na) rmse(NaRFPred, RFSalts_Test$Na)

#For F FRFPred<- predict(RFModelF, newdata = RFSalts_Test) cor.test(FRFPred, RFSalts_Test$F) plot(FRFPred, RFSalts_Test$F) sma(FRFPred ~ RFSalts_Test$F, slope.test=1)

245

mae(FRFPred, RFSalts_Test$F) rmse(FRFPred, RFSalts_Test$F)

#For Sr SrRFPred<- predict(RFModelSr, newdata = RFSalts_Test) cor.test(SrRFPred, RFSalts_Test$Sr) plot(SrRFPred, RFSalts_Test$Sr) sma(SrRFPred ~ RFSalts_Test$Sr, slope.test=1) mae(SrRFPred, RFSalts_Test$Sr) rmse(SrRFPred, RFSalts_Test$Sr)

#For NH3 NH3RFPred<- predict(RFModelNH3, newdata = RFSalts_Test) cor.test(NH3RFPred, RFSalts_Test$NH3) plot(NH3RFPred, RFSalts_Test$NH3) sma(NH3RFPred ~ RFSalts_Test$NH3, slope.test=1) mae(NH3RFPred, RFSalts_Test$NH3) rmse(NH3RFPred, RFSalts_Test$NH3)

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

#Running the Darwin Glacier region test dataset on the generated multiple linear regression models

#For Total salts #Inputting the new test dataset and storing the predicted concentration results Darwin_TSLMPred <- predict(TSreg, newdata = Darwin) #Running a correlation between the measured and predicted values cor.test(Darwin_TSLMPred, Darwin$Total_salt) #Quick visual check of the data plot(Darwin_TSLMPred, Darwin$Total_salt) #Determining the slope by fitting bivariate lines using the standardized major axis (SMA) for measured and predicted values #A slope of 1 is ideal, where there is a direct 1:1 relationship #sma will test the slope estimate and determine the test statistic for the null hypothesis sma(Darwin_TSLMPred~ Darwin$Total_salt, slope.test=1) #Determing the mean absolute error (MAE) and root mean squared error (RMSE)

246

mae(Darwin_TSLMPred, Darwin$Total_salt) rmse(Darwin_TSLMPred, Darwin$Total_salt)

#For N:P ratio Darwin_NPLMPred <- predict(NPreg, newdata = Darwin) cor.test(Darwin_NPLMPred, Darwin$NP) plot(Darwin_NPLMPred, Darwin$NP) sma(Darwin_NPLMPred~ Darwin$NP, slope.test=1) mae(Darwin_NPLMPred, Darwin$NP) rmse(Darwin_NPLMPred, Darwin$NP)

#For Cl Darwin_ClLMPred <- predict(Clreg, newdata = Darwin) cor.test(Darwin_ClLMPred, Darwin$Cl) plot(Darwin_ClLMPred, Darwin$Cl) sma(Darwin_ClLMPred~ Darwin$Cl, slope.test=1) mae(Darwin_ClLMPred, Darwin$Cl) rmse(Darwin_ClLMPred, Darwin$Cl)

#For NO3 Darwin_NO3LMPred <- predict(NO3reg, newdata = Darwin) cor.test(Darwin_NO3LMPred, Darwin$NO3) plot(Darwin_NO3LMPred, Darwin$NO3) sma(Darwin_NO3LMPred~ Darwin$NO3, slope.test=1) mae(Darwin_NO3LMPred, Darwin$NO3) rmse(Darwin_NO3LMPred, Darwin$NO3)

#For Ca Darwin_CaLMPred <- predict(Careg, newdata = Darwin) cor.test(Darwin_CaLMPred, Darwin$Ca) plot(Darwin_CaLMPred, Darwin$Ca) sma(Darwin_CaLMPred~ Darwin$Ca, slope.test=1) mae(Darwin_CaLMPred, Darwin$Ca) rmse(Darwin_CaLMPred, Darwin$Ca)

#For Mg Darwin_MgLMPred <- predict(Mgreg, newdata = Darwin) cor.test(Darwin_MgLMPred, Darwin$Mg)

247

plot(Darwin_MgLMPred, Darwin$Mg) sma(Darwin_MgLMPred~ Darwin$Mg, slope.test=1) mae(Darwin_MgLMPred, Darwin$Mg) rmse(Darwin_MgLMPred, Darwin$Mg)

#For Na Darwin_NaLMPred <- predict(Nareg, newdata = Darwin) cor.test(Darwin_NaLMPred, Darwin$Na) plot(Darwin_NaLMPred, Darwin$Na) sma(Darwin_NaLMPred~ Darwin$Na, slope.test=1) mae(Darwin_NaLMPred, Darwin$Na) rmse(Darwin_NaLMPred, Darwin$Na)

#For K Darwin_KLMPred <- predict(Kreg, newdata = Darwin) cor.test(Darwin_KLMPred, Darwin$K) plot(Darwin_KLMPred, Darwin$K) sma(Darwin_KLMPred~ Darwin$K, slope.test=1) mae(Darwin_KLMPred, Darwin$K) rmse(Darwin_KLMPred, Darwin$K)

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

#Running the Darwin Glacier region test dataset on the generated random forest models

#For Total salts #Inputting the new test dataset and storing the predicted concentration results Darwin_TSRFPred<-predict(RFModelTS, newdata = Darwin) #Running a correlation between the measured and predicted values cor.test(Darwin_TSRFPred, Darwin$Total_salt) #Quick visual check of the data plot(Darwin_TSRFPred, Darwin$Total_salt) #Determining the slope by fitting bivariate lines using the standardized major axis (SMA) for measured and predicted values #A slope of 1 is ideal, where there is a direct 1:1 relationship #sma will test the slope estimate and determine the test statistic for the null hypothesis sma(Darwin_TSRFPred~ Darwin$Total_salt, slope.test=1)

248

#Determing the mean absolute error (MAE) and root mean squared error (RMSE) mae(Darwin_TSRFPred, Darwin$Total_salt) rmse(Darwin_TSRFPred, Darwin$Total_salt)

#For N:P ratio Darwin_NPRFPred<-predict(RFModelNP, newdata = Darwin) cor.test(Darwin_NPRFPred, Darwin$NP) plot(Darwin_NPRFPred, Darwin$NP) sma(Darwin_NPRFPred~ Darwin$NP, slope.test=1) mae(Darwin_NPRFPred, Darwin$NP) rmse(Darwin_NPRFPred, Darwin$NP)

#For Cl Darwin_ClRFPred<-predict(RFModelCl, newdata = Darwin) cor.test(Darwin_ClRFPred, Darwin$Cl) plot(Darwin_ClRFPred, Darwin$Cl) sma(Darwin_ClRFPred~ Darwin$Cl, slope.test=1) mae(Darwin_ClRFPred, Darwin$Cl) rmse(Darwin_ClRFPred, Darwin$Cl)

#For NO3 Darwin_NO3RFPred<-predict(RFModelNO3, newdata = Darwin) cor.test(Darwin_NO3RFPred, Darwin$NO3) plot(Darwin_NO3RFPred, Darwin$NO3) sma(Darwin_NO3RFPred~ Darwin$NO3, slope.test=1) mae(Darwin_NO3RFPred, Darwin$NO3) rmse(Darwin_NO3RFPred, Darwin$NO3)

#For Ca Darwin_CaRFPred<-predict(RFModelCa, newdata = Darwin) cor.test(Darwin_CaRFPred, Darwin$Ca) plot(Darwin_CaRFPred, Darwin$Ca) sma(Darwin_CaRFPred~ Darwin$Ca, slope.test=1) mae(Darwin_CaRFPred, Darwin$Ca) rmse(Darwin_CaRFPred, Darwin$Ca)

#For Mg Darwin_MgRFPred<-predict(RFModelMg, newdata = Darwin)

249

cor.test(Darwin_MgRFPred, Darwin$Mg) plot(Darwin_MgRFPred, Darwin$Mg) sma(Darwin_MgRFPred~ Darwin$Mg, slope.test=1) mae(Darwin_MgRFPred, Darwin$Mg) rmse(Darwin_MgRFPred, Darwin$Mg)

#For Na Darwin_NaRFPred<-predict(RFModelNa, newdata = Darwin) cor.test(Darwin_NaRFPred, Darwin$Na) plot(Darwin_NaRFPred, Darwin$Na) sma(Darwin_NaRFPred~ Darwin$Na, slope.test=1) mae(Darwin_NaRFPred, Darwin$Na) rmse(Darwin_NaRFPred, Darwin$Na)

#For K Darwin_KRFPred<-predict(RFModelK, newdata = Darwin) cor.test(Darwin_KRFPred, Darwin$K) plot(Darwin_KRFPred, Darwin$K) sma(Darwin_KRFPred~ Darwin$K, slope.test=1) mae(Darwin_KRFPred, Darwin$K) rmse(Darwin_KRFPred, Darwin$K)

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

#Exporting Shackleton and Darwin Glacier region RF test results Shackleton_RFtest_results<- data.frame(TaRFPred, NPRFPred, ClO4RFPred, ClRFPred, FRFPred, NO3RFPred, SO4RFPred, PO4RFPred, CaRFPred, MgRFPred, NaRFPred, KRFPred, SrRFPred, NH3RFPred) write.csv(Shackleton_RFtest_results, file = "Shackleton_RFtest_results")

Darwin_RFtest_results <-data.frame(Darwin_TSRFPred,Darwin_NPRFPred, Darwin_ClRFPred,Darwin_NO3RFPred, Darwin_CaRFPred, Darwin_MgRFPred,Darwin_NaRFPred,Darwin_KRFPred) write.csv(Darwin_RFtest_results, file = "Darwin_RF_test_results.csv")

250