University of Southern California Dissertations and Theses

Chemical weathering across spatial and temporal scales: from laboratory experiments to global models

(USC Thesis Other)

Chemical weathering across spatial and temporal scales: from laboratory experiments to global models

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content CHEMICAL WEATHERING ACROSS SPATIAL AND TEMPORAL SCALES: FROM LABORATORY EXPERIMENTS TO GLOBAL MODELS by Mark Albert Torres A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (GEOLOGICAL SCIENCES) August 2015 Copyright 2015 Mark Albert Torres Epigraph It is quite rare that scientists are asked to meet with artists and are chal- lenged to match the others’ creativeness. Such an experience may well humble the scientist. The medium in which he works does not lend itself to the delight of the listener’s ear. When he designs his experiments or executes them with devoted attention to the details he may say to himself. "This is my composition; the pipette is my clarinet". And the orchestra may include instruments of the most subtle design. To others, however, his music is as silent as the music of the spheres. He may say to himself, "My story is an everlasting possession, not a prize composition which is heard and forgotten", but he fools only himself. The books of the great scientists are gathering dust on the shelves of learned libraries. And rightly so. The scientist addresses an infinitesimal audience of fellow composers. His message is not devoid of universality but it’s universality is disembodied and anonymous. While the artist’s communication is linked forever with it’s original form, that of the scientist is modified, amplified, fused with the ideas and results of others, and melts into the stream of knowledge and ideas which forms our culture. The scientist has in common with the artist only this: that he can find no better retreat from the world than his work and also no stronger link with his world than his work. -Max Delbrück in his 1969 Nobel Prize Lecture ii Acknowledgments thanks everybody. iii Contents Epigraph ii Acknowledgments iii List of Tables viii List of Figures ix Abstract xi 1 Introduction 1 1.1 Microbial nutrient acquisition . . . . . . . . . . . . . . . . . . . . . 2 1.2 The hydrologic control on chemical weathering . . . . . . . . . . . . 3 1.3 The acid budget of chemical weathering: catchment scale . . . . . . 4 1.4 The acid budget of chemical weathering: global scale . . . . . . . . 4 2 Microbialaccelerationofsilicatemineraldissolutionviasiderophore production 6 2.1 Linking Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.3.1 Olivine sample . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.3.2 Experimental solutions . . . . . . . . . . . . . . . . . . . . . 10 2.3.3 Batch dissolution experiments . . . . . . . . . . . . . . . . . 11 2.3.4 Solute concentration measurements . . . . . . . . . . . . . . 11 2.3.5 Ligand speciation measurements . . . . . . . . . . . . . . . . 12 2.3.6 Calculation of dissolution rates . . . . . . . . . . . . . . . . 12 2.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 2.4.1 Maximum concentrations and solution saturation states . . . 13 2.4.2 Stoichiometry of dissolution . . . . . . . . . . . . . . . . . . 14 2.4.3 Linearity of solute release with time . . . . . . . . . . . . . . 15 2.4.4 Dissolution rates . . . . . . . . . . . . . . . . . . . . . . . . 15 2.4.5 Spectrophotometric measurements . . . . . . . . . . . . . . . 17 iv 2.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 2.5.1 Non-linearsolutereleaseandthemechanismofdeferoxamine- promoted olivine dissolution . . . . . . . . . . . . . . . . . . 18 2.5.2 Implications for microbially-accelerated silicate mineral dis- solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.5.3 Feedback Model . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3 Geomorphicregimemodulateshydrologiccontrolofchemicalweath- ering in the Andes-Amazon 28 3.1 Linking Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 3.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 3.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 3.3.1 Study Site . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 3.3.2 Sampling Methodology . . . . . . . . . . . . . . . . . . . . . 36 3.3.3 Analytical Methodology . . . . . . . . . . . . . . . . . . . . 37 3.3.4 Data Processing . . . . . . . . . . . . . . . . . . . . . . . . . 38 3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 3.4.1 Atmospheric Inputs . . . . . . . . . . . . . . . . . . . . . . . 40 3.4.2 Intra-site variation in elemental concentrations with runoff . 42 3.4.3 Variations in elemental ratios with runoff . . . . . . . . . . . 46 3.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 3.5.1 Lithologic controls on concentration-runoff and ratio-runoff relationships . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 3.5.2 The effects of secondary mineral formation on concentration- runoff and ratio-runoff relationships . . . . . . . . . . . . . . 52 3.5.3 The effects of changes in fluid flow paths on concentration- runoff and ratio-runoff relationships . . . . . . . . . . . . . . 54 3.5.4 Geomorphic control on concentration-runoff relationships in the Andes Amazon system . . . . . . . . . . . . . . . . . . . 57 3.5.5 ImplicationsfortheroleoffloodplainweatheringintheAma- zon system . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 3.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 4 Geomorphic influence on the acid budget of weathering in the Andes-Amazon system 70 4.1 Linking Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 4.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 4.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 4.3.1 Study Site . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 4.3.2 Solid phase geochemical analyses . . . . . . . . . . . . . . . 74 4.3.3 Dissolved phase geochemical analyses . . . . . . . . . . . . . 74 v 4.3.4 Tributary mixing model . . . . . . . . . . . . . . . . . . . . 75 4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 4.4.1 Solid phase analyses . . . . . . . . . . . . . . . . . . . . . . 75 4.4.2 Dissolved phase analyses . . . . . . . . . . . . . . . . . . . . 78 4.4.3 Tributary mixing . . . . . . . . . . . . . . . . . . . . . . . . 84 4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 4.5.1 Lithologic source partitioning - qualitative . . . . . . . . . . 87 4.5.2 Lithologic source partitioning - quantitative . . . . . . . . . 89 4.5.3 Sulfate sources . . . . . . . . . . . . . . . . . . . . . . . . . 91 4.5.4 Tributary mixing . . . . . . . . . . . . . . . . . . . . . . . . 94 4.5.5 The acid budget of chemical weathering . . . . . . . . . . . 98 4.5.6 Riverine DIC . . . . . . . . . . . . . . . . . . . . . . . . . . 99 4.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 5 Sulphide oxidation and carbonate dissolution as a source of CO 2 over geological timescales 105 5.1 Linking Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 5.2 Introduction and Main Text . . . . . . . . . . . . . . . . . . . . . . 105 5.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 5.3.1 Calculation of CO 2 Fluxes from Modern Rivers . . . . . . . 112 5.3.2 Conceptual Box Model . . . . . . . . . . . . . . . . . . . . . 114 5.3.3 Li and Elderfield Mass Balance Model . . . . . . . . . . . . 114 5.3.4 Sulfur Isotope Mass Balance Model . . . . . . . . . . . . . . 117 5.4 Supplementary Results . . . . . . . . . . . . . . . . . . . . . . . . . 119 5.4.1 Sulfur Cycle Mass Balance Model . . . . . . . . . . . . . . . 119 5.4.2 Determining the effect of sulfide oxidation on Cenozoic pO 2 123 6 Conlusions 133 6.1 Microbe-mineral interactions . . . . . . . . . . . . . . . . . . . . . . 133 6.2 Hydrologic controls on chemical weathering . . . . . . . . . . . . . . 134 6.3 Sulfide oxidation in the Andes/Amazon . . . . . . . . . . . . . . . . 134 6.4 The Cenozoic isotope-weathering paradox . . . . . . . . . . . . . . 134 Reference List 136 A Mineral Dissolution Model 154 B Feedback Model 156 C Atmospheric input correction 159 vi D Detailed methods 161 D.0.1 Solid phase geochemical analyses . . . . . . . . . . . . . . . 161 D.0.2 Dissolved phase geochemical analyses . . . . . . . . . . . . . 164 E Tributary mixing model 169 F Lithologic source mixing model 171 vii List of Tables 2.1 Calculated Linear Dissolution Rates . . . . . . . . . . . . . . . . . . 27 3.1 Catchment Metadata . . . . . . . . . . . . . . . . . . . . . . . . . . 36 3.2 Solute power-law exponents . . . . . . . . . . . . . . . . . . . . . . 68 4.1 Calculated water discharges based on conservative tracers . . . . . . 102 5.1 Large river CO 2 fluxes . . . . . . . . . . . . . . . . . . . . . . . . . 114 viii List of Figures 2.1 Dissolution stoichiometry . . . . . . . . . . . . . . . . . . . . . . . . 15 2.2 Solute release with time . . . . . . . . . . . . . . . . . . . . . . . . 16 2.3 Effect of deferoxamine B on olivine dissolution rates . . . . . . . . . 17 2.4 Ligand-metal complex absorption measurements . . . . . . . . . . . 18 2.5 Model of the effects of surface coatings on dissolution rates . . . . . 21 2.6 Fe-dependent microbial growth and ligand production . . . . . . . . 24 2.7 Siderophore production feedback model . . . . . . . . . . . . . . . . 26 3.1 Map of the study site. . . . . . . . . . . . . . . . . . . . . . . . . . 33 3.2 Example of the atmospheric input correction method. . . . . . . . . 39 3.3 Wayqecha catchment (Way) concentration-runoff relationships. . . . 43 3.4 San Pedro catchment (SP) concentration-runoff relationships. . . . 44 3.5 Manu Learning Center catchment (MLC) concentration-runoff rela- tionships. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 3.6 CICRA catchment concentration-runoff relationships. . . . . . . . . 47 3.7 Wayqecha catchment (Way) ratio-runoff relationships. . . . . . . . . 48 3.8 San Pedro catchment (SP) ratio-runoff relationships. . . . . . . . . 49 3.9 Manu Learning Center catchment (MLC) ratio-runoff relationships. 50 3.10 CICRA catchment ratio-runoff relationships. . . . . . . . . . . . . . 51 3.11 Geochemical behavior of Li . . . . . . . . . . . . . . . . . . . . . . . 53 3.12 Schematic of the hypothesized effects of flow path variability on Andean concentration-runoff relationships. . . . . . . . . . . . . . . 56 3.13 Correlation between the calculated power law exponents and mean catchment slope angle for all solutes. . . . . . . . . . . . . . . . . . 58 ix 3.14 Correlation between the calculated power law exponents and mean catchment slope angle for Na and Si. . . . . . . . . . . . . . . . . . 59 3.15 Schematic of the coupled variation in fluid transit times and mineral residence times and their effects on concentration-runoff relation- ships across the mountain-to-floodplain transition. . . . . . . . . . . 69 4.1 Tributary Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 4.2 Solid phase geochemistry . . . . . . . . . . . . . . . . . . . . . . . . 78 4.3 Dissolved phase geochemistry . . . . . . . . . . . . . . . . . . . . . 80 4.4 Radiogenic strontium isotope geochemistry . . . . . . . . . . . . . . 81 4.5 Sulfur isotope geochemistry . . . . . . . . . . . . . . . . . . . . . . 83 4.6 Carbon isotope geochemistry . . . . . . . . . . . . . . . . . . . . . . 84 4.7 Downstream changes in water isotope ratios . . . . . . . . . . . . . 86 4.8 Water discharges calculated based on water isotope mixing . . . . . 87 4.9 Lithologic mixing model . . . . . . . . . . . . . . . . . . . . . . . . 91 4.10 S isotope fractionation at geomorphic steady state . . . . . . . . . . 95 4.11 Conservative tributary mixing: Wet season . . . . . . . . . . . . . . 96 4.12 Conservative tributary mixing: Dry season . . . . . . . . . . . . . . 103 4.13 The acid budget of weathering . . . . . . . . . . . . . . . . . . . . . 104 5.1 CO 2 fluxes in modern rivers . . . . . . . . . . . . . . . . . . . . . . 109 5.2 Conceptual box model . . . . . . . . . . . . . . . . . . . . . . . . . 126 5.3 Isotope mass balance model . . . . . . . . . . . . . . . . . . . . . . 127 5.4 Oxygen consumption and carbon dioxide release . . . . . . . . . . . 128 5.5 Modeled Cenozoic pyrite burial fluxes . . . . . . . . . . . . . . . . . 129 5.6 Modeled Cenozoic Δ pyrite−seawater . . . . . . . . . . . . . . . . . . . 130 5.7 Modeled Cenozoic δ 34 S riverine input values . . . . . . . . . . . . . 131 5.8 Implications of S cycle models for Cenozoic pO 2 . . . . . . . . . . . 132 D.1 Correction for silicate contamination of leachates . . . . . . . . . . 168 x Abstract When exposed at Earth’s surface, rocks are out of thermodynamic equilib- rium with respect to their environment. This disequilibrium drives the chemical transformation, or weathering, of these rocks into soils and governs the chemi- cal composition of natural waters and the atmosphere. Over geologic timescales, complex feedbacks associated with weathering processes are presumed to regulate the concentrations of CO 2 and O 2 in the atmosphere with profound implications for the habitability of the planet. However, a mechanistic understanding of how biologic, tectonic, and climatic conditions interact to control weathering fluxes has remainedelusive. Inpart, ourunderstandingofweatheringprocessesishinderedby the fact that they operate continuously over an enormous range of spatial (atomic to global) and temporal (microseconds to millions of years) scales, but we can only make measurements over discrete ranges of these values. While each scale of observation available offers unique insights, it is often difficult to link observations made at different scales. For my Ph.D., I focused on three distinct projects that span the range of observable scales in order to better understand the links between chemical weathering and long-term biogeochemical cycles. Chapter 2: Laboratory insights into microbial mineral dissolution Rocksandmineralsrepresentamajorreservoirofbio-essentialnutrients. While abundant, someoftheselithogenicnutrients, likeiron, arenotreadilybio-available. As a result, many organisms produce metal-binding ligands to scavenge these trace nutrients from the environment. Using targeted laboratory experiments with live microbial cultures and purified microbial ligands, I explored efficacy by which microbes can access trace nutrients from common silicate minerals (Torres et al., in prep). In addition to providing insight into biological nutrient scavenging strate- gies, this work also provides the basic research necessary to develop microbe-based xi CO 2 sequestration techniques since the dissolution of silicate minerals for nutrient acquisition also sequesters CO 2 . Chapters 3 & 4: Geomorphic control on the hydrology and carbon budget of weathering Erosional processes and hydrology are known to influence chemical weathering rates by controlling the timescales over which minerals react. Accurately describ- ing the complex linkages between weathering, erosion, and hydrology observed in natural environments remains a major research challenge. To help address this problem, a major part of my Ph.D. was focused on characterizing how chemical weathering and hydrology are coupled in distinct erosional environments. This work combines hydrologic monitoring, solute chemistry, and water isotope analy- ses in order to robustly document how water is stored in catchments and link this to measured solute fluxes from chemical weathering (Clark et al., 2014; Torres et al., 2015). The results of this study are intriguing in that the hydrological control of weathering was found to vary predictably with the erosional regime, which has importantimplicationsforhowchangesintectonicactivityaffectglobalweathering fluxes. By affecting the timescales over which weathering reactions occur, erosional processes also influence which minerals react due to the intrinsic variability in the reaction rates of different minerals. Rapid erosion rates favor the oxidation of sulfide minerals relative to the dissolution of silicate minerals, which leads to the release of CO 2 into the ocean-atmosphere system. To trace sulfide oxidation and its effect on the carbon budget, I combined multiple isotopic systems (e.g., S, C, and Sr) with major and trace element analyses (Torres et al., in prep). By making observations in catchments with diverse erosional regimes, it was possible to inter- rogate how sulfide oxidation fluxes relate to erosional processes. My results showed that sulfide oxidation dominates in rapidly eroding environments and leads to the significant release of CO 2 . This is in contrast to more slowly eroding environments, where CO 2 consumption during silicate weathering dominates. Chapter 5: The evolution of the Cenozoic carbon cycle My research on sulfide oxidation in modern systems suggested a link between tectonicupliftandthecarbonbudgetofweatheringprocesseswithimportantimpli- cations for the long-term carbon cycle. To test this hypothesis, I incorporated the xii effects of sulfide-oxidation driven CO 2 release into a model of the Cenozoic car- bon cycle. The Cenozoic carbon cycle has long plagued geochemists as isotopic records suggest changes in weathering fluxes that appear to be inconsistent with the requirement of mass balance in the long-term carbon cycle. By incorporating sulfide oxidation as a CO 2 source, I was able to provide a novel solution to the "Cenozoic isotope-weathering paradox" (Torres et al., 2014). xiii Chapter 1 Introduction As conscious habitants of the Earth, we have a strong desire to understand how ourplanet"works". Whatcontrolsthearabilityofsoilsandoverwhattimescalesare nutrientsdepleted? HowlongiswaterstoredwithintheEarthandhowdoreactions betweenwaterandrockmodifyitschemistry? Whyisourplanethabitableandhow are equable conditions maintained over geologic timescales? The list of questions we have goes on. Addressing these questions is the objective of the Earth Sciences. At its most practical, the scientific study of the Earth offers insights, tools, and models that can guide society in its use and, hopefully, stewardship of the planet. However, the Earth does not readily reveal the nature of its inner workings. All key processes operate continuously over an enormous range of spatial and temporal scales of which most are out of our range of direct observation. As Earth Scientists, our challenge is to infer from our limited vantage point and then extrapolate. In this Ph.D. dissertation, I report the results of three projects conducted from separate vantage points using the tools of analog experiments, field measurements, and mathematical models. Each vantage point/tool offers its own unique insights into how a particular process operates over a discrete range of time and space. By linking observations made at different scales, greater insight into a particular problem is gained. With analog experiments, the Earth scientist defines nearly every aspect of the system, allowing for unparalleled insight into its inner workings. Addition- ally, the experimenter can rapidly change direction as more data is collected and interpreted. The challenge is, however, to extrapolate the information gleaned from within a laboratory to the world at large. Often, we do not know exactly how the Earth has been constructed, making it difficult to accurately construct the Earth within our laboratory. Similarly, many "experiments" conducted by the Earth occur on spatial and temporal scales much too large for proper laboratory study. 1 To circumvent these and the other many problems with analog experiments, many scientists elect to analyze the results of Earth’s own "experiments" by con- ducting fieldwork. Field studies ensure that system was defined at an appropriate scaleforthescientificquestion. However, withfieldexperiments, weoftentodonot have the ability or means to measure every parameter necessary to fully constrain a system. No matter where the inquiry starts, be it in the lab or in the field or both, mathematical models are an effective way to synthesize results. In part, this is because models allow access to spatial and temporal scales otherwise inaccessible to humans. The challenge is to know what processes are necessary to include within a model and how to properly parameterize them. In this Ph.D. dissertation, the aforementioned tools are applied to further our understanding chemical weathering and Earth’s long-term biogeochemical cycles. When exposed at Earth’s surface, igneous, metamorphic, and sedimentary rocks are out of thermodynamic equilibrium with respect to their environment. This dis- equilibrium drives the chemical transformation, or weathering, of these rocks into soilsandgovernsthechemicalcompositionofnaturalwatersandtheatmosphereby releasing soluble elements and consuming reactive gases. Over geologic timescales, complex feedbacks associated with weathering processes are presumed to regulate the concentrations of CO 2 and O 2 in the atmosphere with profound implications for the habitability of the planet. However, a mechanistic understanding of how biologic, tectonic, and climatic conditions interact to control weathering rates has remained elusive. This Ph.D. dissertation aims to address key knowledge gaps in our current understanding of weathering systems by focusing on 1) microbial nutrient acquisition from silicate minerals, 2) the hydrologic control on chemical weathering in mountains and floodplains, 3) the importance of trace sulfide min- erals at the watershed scale, and 4) the importance of trace sulfide minerals at the global scale. 1.1 Microbial nutrient acquisition Due to their abundance at Earth’s surface, silicate minerals represent an impor- tant reservoir of essential nutrient elements required for biological activity. Despite 2 their abundance, many silicate-bound nutrients are not readily "bio-available" due to the slow rate at which silicate minerals dissolve and the rapid precipitation of secondary metal-bearing mineral phases. To acquire nutrients from the envi- ronment, many organisms produce organic metal-binding ligands that accelerate mineral dissolution rates and stabilize nutrients as dissolved organic complexes (Page & Huyer, 1984; Römheld & Marschner, 1986; Liermann et al., 2000). In chapter 2, the effects of organic ligands on the kinetics of Fe release from olivine are measured and used to infer dissolution mechanisms and explore the feedbacks that may control biologic nutrient acquisition in natural systems. 1.2 The hydrologic control on chemical weather- ing Considering the slow reaction kinetics of minerals at Earth’s surface and their heterogeneous distribution, it is readily apparent that chemical weathering pro- cesses should be extremely sensitive to the supply, routing, and transport of water. However, many of the details of the hydrologic controls on chemical weathering in natural systems are not completely understood. This reflects, in part, knowledge gaps in both catchment hydrology and chemical weathering. For example, the timescale over which water is stored during flow through catchment systems is poorly constrained and likely to vary spatially and temporally (Kirchner, 2015). Similarly, forpoorlyunderstoodreasons, mineralreactionskineticsmeasuredinthe laboratory are orders of magnitude faster than those measured in the field. Since these knowledge gaps may be related (e.g., Maher 2010, 2011; Reeves & Roth- man 2013), the investigation of weathering processes over a range of hydrologic conditions may provide keys insights into both catchment hydrology and chemical weathering. In chapter 3, the results of a field study on the relationships between hydrology and chemical weathering in the rapidly eroding Peruvian Andes and the depo- sitional Amazonian foreland floodplain are reported. Since erosion, along with hydrology, is thought to be an important control on chemical weathering (e.g., West 2012; Maher & Chamberlain 2014), the design of this study allows for novel insight into the different controls on chemical weathering natural systems. The 3 results suggest a coupling between hydrologic and erosional processes that ampli- fies differences in chemical weathering fluxes between rapidly- and slowly-eroding environments. 1.3 The acid budget of chemical weathering: catchment scale Both hydrologic and erosional processes set the characteristic timescales over which chemical weathering reactions occur in surface environments (e.g., West 2012; Maher & Chamberlain 2014). In addition to controlling the mass fluxes of weathering products, which is described in chapter 3, variations in the timescale over which weathering reactions occur should also control the extent to which trace reactive phases dominate weathering budgets. One such trace reactive phase, pyrite (FeS 2 ), is particularly relevant since it is a common mineral and, when exposed to O 2 , reacts to form sulfuric acid (H 2 SO 4 ). While the behavior of sulfuric acid in surface environments is complex, its effects on acid-base chemistry can be linked with the release of CO 2 stored in rocks into the ocean/atmosphere system (Calmels et al., 2007; Torres et al., 2014). Additionally, the oxidation of pyrite and production of sulfuric acid acts as a positive feedback on chemical weathering by increasing the porosity and permeability of the host rock (Brantley et al., 2013). In chapter 4, the results of a multitude of geochemical analyses of samples from the Andes-Amazon field sites are presented and used to quantify the importance of sulfuric acid weathering across a natural gradient in weathering duration. These results are supplemented with a budgeting analysis of water and solute fluxes that provides insights into how chemical signals originating from the Andes Mountains are modified as the rivers flow through the large foreland-floodplain region. 1.4 The acid budget of chemical weathering: global scale The results of the field study presented in chapter 4 suggest that mountainous regions act as sources of atmospheric CO 2 and sinks of atmospheric O 2 due to the 4 preferential reaction of trace sulfide minerals. If generally applicable, this result has important implications for how orogenic events in Earth’s history affect the long-term C and O 2 cycles. In chapter 5, the effects of enhanced sulfide oxidation in mountainous regions are considered in a model of Cenozoic uplift and C cycle perturbations. During the Cenozoic, many major orogenies (e.g., the Himalayan and Andean orogenies) are thought to have profoundly affected Earth’s carbon cycle by accel- erating the rate of CO 2 drawdown by enhancing the supply and weathering of silicate minerals in surface environments (Raymo et al., 1988). While consistent with our understanding of the controls on silicate weathering in modern systems (Gaillardet et al., 1999; West, 2012) and marine isotopic proxies sensitive to the effects of silicate weathering (McArthur et al., 2001; Klemm et al., 2005; Misra & Froelich, 2012), the hypothesis of tectonically-enhanced chemical weathering is inconsistent with the need for mass-balance in the long-term C cycle (Berner & Caldeira, 1997). By adding the effects of tectonically-enhanced sulfide oxidation, theresultsofchapter5provideanovelsolutiontotheCenozoic"isotope-weathering paradox". 5 Chapter 2 Microbial acceleration of silicate mineral dissolution via siderophore production 2.1 Linking Statement This chapter is focused around laboratory mineral dissolution experiments, which offer insights into weathering processes as they occur at small spatial and temporal scales. While the experimental procedure itself is conceptually simple and relatively easy to conduct (i.e. putting rocks in a bottle), the results of lab- oratory dissolution experiments are notoriously complex. For example, the intra- laboratory reproducibility of a measured dissolution rate is typically on the order of 30% (Rimstidt et al., 2012). For the same mineral under the same environmen- tal conditions (e.g., pH, temperature, and fluid composition), comparisons show that inter-laboratory variability is on the order of 300% (Rimstidt et al., 2012). This implies that there are additional unknown sources of variance, including het- erogeneity at the mineral surface (e.g., Fischer et al. 2012). Thus, when we try to understand weathering at larger spatial and tempo- ral scales, which is the focus of subsequent chapters in the dissertation, we are attempting to scale a process that we do not completely understand. So, while this chapter may appear thematically distinct from the rest of this dissertation, its presence is defensible in that progress in understanding weathering processes at the catchment scale and beyond requires developments in our understanding of the basic chemistry and mechanisms of mineral dissolution, which can only be gained through laboratory experiments. Here, the problem is framed around understand- ing microbial nutrient acquisition, but the results provide new and fundamental insights into how the chemical evolution of mineral surfaces affects reaction rates. 6 Chapter 2, like all other chapters in this dissertation, is the product of collab- oration. Drs. Ken Nealson and Joshua West are acknowledged for their insights and supervision. Sijia Dong is responsible for conducting the tests of the pH buffer and the set of 340-hour dissolution experiments that proved to be extraordinar- ily useful. David Hercules, Natalie DeVries, Kevin DiBella, and Daisy Arriaga are acknowledged for their summer work, which proved useful for optimizing the experimental protocol. Pratixa Savalia, Andrea Cheung, and Dr. Sarah Bennett are all thanked for their help in the laboratory during the early stages of this project. Support for this work was provided by a C-DEBI graduate fellowship, an SEPM graduate student grant, an IAGC graduate student grant, and a NSF grant awarded to Drs. Joshua West and Ken Nealson. 2.2 Introduction Silicatemineralsrepresentanimportantreservoirofinorganicnutrientelements (e.g., Ca, K, and Fe) that are required for biologic activity. These "lithogenic" nutrients can limit biologic activity when supplied to the system in insufficient quantities (Vitousek et al., 2003; Hahm et al., 2014) or if the nutrient element is "trapped" within refractory solid phases (Calvaruso et al., 2006; Bellenger et al., 2008; Mendez et al., 2010). The latter is particularly true in the case of Fe, which, due to the low solubility of many Fe-bearing mineral phases, is typically present at low concentrations in most natural waters (Stumm & Lee, 1961; Hem, 1972; White et al., 1985). For most organisms, Fe is an essential nutrient since it is involved in many basic cellular reduction/oxidation (redox) reactions. Iron is also a co-factor in many enzymes required for other important biologic processes, including many that are important for the cycling of nitrogen (Berges & Mulholland, 2008). In addition to being a nutrient, Fe-bearing compounds can also serve as either an electron donor or acceptor for some microorganisms depending upon its oxidation state (Bird et al., 2011). In most primary silicate minerals, Fe 2+ is the dominant Fe redox species. At circum-neutral pH, Fe 2+ is rapidly oxidized, even abiotically, by O 2 yielding Fe 3+ (Stumm & Lee, 1961; Hem, 1972; White et al., 1985). Due to the low solubilities of 7 common ferric oxyhydroxide minerals and rapid oxidation kinetics, Fe 2+ oxidation resultsinverylowaqueousFeconcentrationsinnaturaloxygenatedwaters(Stumm & Lee, 1961; Hem, 1972; White et al., 1985). Under more reducing conditions where Fe 2+ oxidation is inhibited, the solubilities of Fe 2+ -sulfide, Fe 2+ -carbonate, Fe 2+ -phosphate, and Fe 2+ -silicate minerals regulate aqueous Fe concentrations. To cope with low concentrations of aqueous Fe, many organisms are capable of producing organic ligands, termed siderophores, that dissolve particulate Fe phases and stabilize dissolved Fe 3+ (Römheld & Marschner, 1986; Page & Huyer, 1984;Liermannetal.,2000). Theseorganicligandsarestructurallydiverse, butare typicallyhexadentateligandsthatselectivelybindFe 3+ withthreebidentatemetal- binding functional groups (Neilands, 1995; Homann et al., 2009; Budzikiewicz, 2010). Commonly, the metal-binding functional groups are catechols, hydroxamic acids, and/or α-hydroxycarboxylic acids (Neilands, 1995; Homann et al., 2009; Budzikiewicz, 2010). Siderophores are ubiquitous in the environment (Powell et al., 1980; Holmström & Lundström, 2004; Essén et al., 2006; Vraspir & Butler, 2009), which suggests that they play an important role in the cycling of nutrient Fe in natural systems. While much research has focused on the structure (Neilands, 1995; Homann et al., 2009; Budzikiewicz, 2010) and production (Fekete et al., 1983; Sweet & Douglas, 1991; Wilhelm et al., 1996; Berraho et al., 1997) of siderophores, there still remain outstanding questions about their reactivity with Fe-bearing silicate mineral phases. Previous studies have shown that the model siderophore deferox- amine B accelerates the dissolution rates of biotite (Bray et al., 2015), hornblende (Liermann et al., 2000), palygorskite (Shirvani & Nourbakhsh, 2010), sepiolite (Shirvani & Nourbakhsh, 2010), and smectite (Haack et al., 2008). Importantly, all of these minerals contain not only Fe but also structural Al 3+ , which is also strongly complexed by siderophores (Evers et al., 1989). The extent to which the acceleration of dissolution rates is due to the complexation of Al versus Fe is poorly constrained, which makes it difficult to extrapolate these results to other Fe-bearing silicates. An additional complication is that organic ligands can affect dissolution rates both by changing the activity of dissolved metals that inhibit dissolution (Oelkers & Schott, 1998) and adsorbing to the mineral surface and weakening bonds (Olsen & Rimstidt, 2008). 8 In order to understand how siderophores react with primary Fe-silicate min- erals, olivine is a useful model mineral since it does not contain structural Al 3+ and its dissolution rate is independent of the aqueous activity of its constituent ions at far-from-equilibrium conditions at acidic to neutral pH values (Pokrovsky & Schott, 2000; Oelkers, 2001). This limits the number of mechanisms by which dissolution rates can be enhanced. Furthermore, olivine is abundant at Earth’s surface and plays an important role in Earth’s long term C cycle by providing a largefluxofMg-alkalinitytotheocean(Dessertetal.,2003)followingtheequation: Mg 2 SiO 4 + 4CO 2 + 4H 2 O↔ 2Mg 2+ + 4HCO − 3 +H 4 SiO 4 (2.1) These factors make research on the dissolution rate of olivine important for under- standing the natural carbon cycle as well as the feasibility of CO 2 sequestra- tion techniques based on accelerated silicate mineral dissolution (e.g., Köhler et al. 2010). In chapter 2, I explore the kinetics and mechanism of biotic Fe- bearing silicate mineral dissolution using olivine ((Mg,Fe) 2 SiO 4 ) and deferoxamine B (N’-5-[Acetyl(hydroxy)amino]pentyl-N-[5-(4-[(5-aminopentyl)(hydroxy)amino]- 4-oxobutanoylamino)pentyl]-N-hydroxysuccinamide) as a model mineral and lig- and respectively. While abiotic mineral dissolution experiments with purified microbial ligands are invaluable for understanding the potential effects of microbial activity on min- eral dissolution rates, extrapolating experimental results to natural systems is not straightforward. In part, this is because the production of ligands is highly regu- lated by the availability of Fe (Fekete et al., 1983; Page & Huyer, 1984; Sweet & Douglas, 1991; Wilhelm et al., 1996; Berraho et al., 1997). This should result in a negative feedback between ligand production and mineral dissolution rates that may control the extent to which biological activity accelerates mineral dissolution reactions in natural systems. 9 2.3 Methods 2.3.1 Olivine sample A natural olivine sample was taken from the University of Southern California mineralogy collection and used for all of the dissolution experiments. The sample consisted of very coarse, sand sized (i.e. 1-2 mm) olivine crystals. Crystals without visible mineral inclusions or surface coatings were hand-picked under a binocular microscope and crushed using either an agate mortar and pestle or a ball mill. The crushed samples were then dry sieved in order to isolate the 75-175 μm grain size fraction. The 75-175μm grain size fraction was then cleaned ultrasonically in de-ionized water (DIW) in order to remove any fine-grained material adhering to the larger crystals. The cleaning procedure was repeated until the DIW remained clear after sonication. The clean 75-175 μm grain size fraction was then dried in an oven overnight at 60 ◦ C. In order to determine its bulk chemical composition, a portion of the olivine sample was ground to a fine powder using a ball mill, doped with ultra high purity quartz, and twice fused with a Li-metaborate flux at 1000 ◦ C in a muffle fur- nace. After the second fusion, the glass bead was polished and analyzed for major and trace element concentrations using a PanAnalytical Axios X-Ray Fluorescence Spectrometer (XRF) at Pomona College. 2.3.2 Experimental solutions A chemical pH buffer was used to maintain a constant pH of approximately 7.5 during the course of the experiments. Of the available pH buffers, TRIS/HCl was selected because 1) TRIS ions do not complex Mg to a significant degree (Hanlon et al., 1966) 2) TRIS ions do not cause a prohibitively large interference with the analytical techniques to measure solute concentrations (see section 2.3.4 below), and 3) the mixture of high purity TRIS (J.T. Baker Ultrapure bioreagent) and teflon-distilled HCl had low background Mg, Fe, and Si concentrations (i.e. below the detection limit; see section 2.3.4 below). A concentration of 10 mM TRIS was used in all of the experiments. Both measurements and speciation calculations confirmed that this buffer concentration was sufficient to keep pH within 0.1 pH units of the starting value for the experimental conditions. 10 Deferoxamine was purchased as a dry mesylate salt (Sigma) and was stored following the supplier’s recommendation. This dried powder was dissolved in DIW to form a concentrated stock solution (i.e. 4.7 mM) that could be mixed with the buffer solution for each experiment. The concentrated Deferoxamine stock was kept in the dark and refrigerated near 1 ◦ C. Titration of the Deferoxamine stock solution with an FeCl 3 /HCl solution was regularly performed in order to ensure that the stock remained stable during storage. 2.3.3 Batch dissolution experiments Batch dissolution experiments were conducted in 125 mL polypropylene (PP) or low-density polyethylene (LDPE) bottles. Attempts were made to perform the experiments in 50 mL PP centrifuge tubes, but these vessels were found to inhibit solution mixing. To each bottle, 0.3± 0.01 g of olivine was added and then mixed with 47± 16 mL of TRIS/HCl buffer and 0-2.7 mL of Deferoxamine stock solution in order to perform experiments across a range of ligand concentrations (0 to 281 μM). After the olivine and experimental solutions were mixed, a sample was immediately taken before the bottle was placed into an incubator/shaker table set to 30 ◦ C and 200 rpm. Subsequent samples were taken at regular intervals (approximately every 4 hours) starting either immediately for some experiments, or after 12 hours of reaction for others. The total mass of the reactor bottle was weighed before and after each sample was taken in order to quantify the volume removed by sampling and volume loss due to evaporation. While most experiments lasted between 12 and 40 hours, a subset of control experiments (i.e. no deferoxamine) were run for up to∼340 hours at 25 ◦ C with samples taken every ∼84 hours. 2.3.4 Solute concentration measurements SamplescollectedfromthebatchdissolutionexperimentswereanalyzedforMg, Si, and Fe concentrations by Microwave Plasma-Atomic Emission Spectroscopy (MP-AES). Prior to analysis, samples were acidified with concentrated distilled HCl (1 μL HCl/ mL sample). Synthetic standards were made by mixing single 11 element standard solutions (from Inorganic Ventures) with the acidified TRIS/HCl in order to match the matrix of the sample solutions. 2.3.5 Ligand speciation measurements In order to identify reaction products, samples were analyzed for their absorbance at wavelengths of 200-650 nm using a Shimazdu UV2600 scanning spectrophotometer. The sample spectra were compared with mixtures of deferox- amine and excesses of either MgCl 2 , Fe 3+ Cl 3 , or silica in a TRIS/HCl matrix. For Fe, titration experiments were also performed in order to determine how the shape of the spectra and the absolute absorbances at 250 and 430 nm changed with Fe 3+ concentrations. 2.3.6 Calculation of dissolution rates For all experiments, the measurements of raw concentration and solution mass were used to calculate the total mass of Mg, Si, and Fe released into solution at eachtimepointtakingintoaccounttheamountthatwasremovedduringsampling. It was observed that Mg was immediately released into solution at the start of each experiment, sothismasswasremovedfromeachtimepoint. Thecalculatedmasses of solutes released were normalized to the mass of olivine used in the experiment and then used to calculate dissolution rates by regressing all of the data against cumulative experiment time. Experimental data were fit using a linear model M =k linear ×t +a (2.2) where M is the mass of solute released per gram of olivine (μmol/g), t is time in hours, k linear is the solute release rate, and a is a constant. In addition to the linear model, the data from the∼340 hour long control experiment were also fit with a power-law model M =k power ×t b +c (2.3) wherek power andc are constants andb is the power law exponent. This formulation assumes that the change in rate with time follows a power-law and thus its inte- gral (i.e. the mass released at each time step) contains an additional constant of 12 integration (i.e. c). I do not include the origin (i.e. 0 time and 0 solute release) in any of the regressions. All regressions were calculated using the MATLAB 2013a curve fitting toolbox and the calculated regression coefficients, their 95% confi- dence intervals, and the R 2 adjusted for the degrees of freedom in the regression equation are reported in Table 2.1. 2.4 Results 2.4.1 Maximum concentrations and solution saturation states For all experiments (i.e. both with and without deferoxamine), the first sample collected, which was taken immediately after the experiment was initiated, con- tained a significant amount of Mg (1.5-16 μM). In the control experiments (no added deferoxamine), measured Mg and Si concentrations reached maximum con- centrations of 43 μM and 24 μM over the course of the approximately 40 hour experiments. At the experimental pH and temperature, PHREEQC calculations using the MINETQ database show that the solution remained undersaturated with respect to all Mg- and Si-bearing mineral phases when these maximum concentra- tions were reached. At 340 hours, the solutions reached maximum Mg and Si concentrations of 87 μM and 52 μM respectively and remained undersaturated with respect to all Mg- and Si-bearing mineral phases. Fe concentrations were below the analytical detection limit at all time points (< 1 μM). In the deferoxamine experiments, measured Mg, Fe, and Si concentrations reached maximum concentrations of 189 μM, 106 μM, and 21 μM over the course of the approximately 40 hour experiments. At the experimental pH and tempera- ture, PHREEQC calculations using the MINETQ database show that the solution remained undersaturated with respect to all Mg- and Si-bearing mineral phases when these maximum concentrations were reached if all of the measured Fe was assumed to be complexed by deferoxamine. 13 2.4.2 Stoichiometry of dissolution The XRF analysis of the olivine sample required doping with ultra high purity quartz. As a result, Si could not be measured reliably. In order to compare the stoichiometryofthesolutemeasurementswiththestoichiometryofthesolidolivine sample, I calculate the Mg/Si and Fe/Si ratios of the olivine sample as Mg/Si olivine = Mg olivine 0.5× (Mg olivine +Fe olivine ) (2.4) and Fe/Si olivine = Fe olivine 0.5× (Mg olivine +Fe olivine ) (2.5) A comparison of measured Mg/Si and Fe/Si ratios to those calculated with equa- tions 2.4 and 2.5 for 28859 olivine samples with Mg, Si, and Fe concentrations mea- surementstakenfromtheGEOROCdatabase(Olivines_1.csvandOlivines_2.csv) reveals that 99% of the analyses have calculated ratios that differ by≤ 10% from the true ratios. For this reason, I assign an uncertainty of 10% for the calculated Mg/Si and Fe/Si ratios of the olivine sample when comparing them to the solute measurements. In the control experiments, Mg/Si ratios calculated for each individual time point match the calculated stoichiometry of the olivine sample (Mg/Si olivine = 1.8 ± 0.18; Figure 2.1a) for all but one of the experiments. The one experiment that had non-stoichiometric Mg/Si ratios showed anomalously low Mg concentrations and is not considered in further analyses. Since Fe concentrations were below the analyticaldetectionlimitforallofthesamplesfromthecontrolexperiments, solute Fe/Si were always less than stoichiometric (Fe/Si olivine = 0.18± 0.02; Figure2.1a). In the deferoxamine B experiments, Mg/Si and Fe/Si ratios calculated for each individual time point show a range of values of which some are consistent with the stoichiometry of the olivine sample Figure 2.1b). The samples with non- stoichiometric ratios may potentially be affected by the dissolution of some trace Mg, Fe, and Si bearing phases. However, the ratios of the Mg and Si release rates calculated for each experiment are consistent with the stoichiometry of the olivine sample (Figure 2.1a). So, while the dissolution of trace phases might affect the measured concentrations at a single time point, they do not significantly affect the measured rates. 14 1.4 1.6 1.8 2 0 0.1 0.2 Mg/Si Fe/Si 0 0.2 0.4 0.6 0 0.1 0.2 0.3 0.4 0.5 Mg release rate ( μMols g −1 hr −1 ) Si release rate ( μMols g −1 hr −1 ) Olivine (Mg = Si x 1.80 ± 0.18) Fit (Mg = Si x 1.75 ± 0.29) Mineral Stoichiometry Control Experiments Deferoxamine (point size = concentration) A B Figure 2.1: Dissolution stoichiometry. Control samples are shown as cyan tri- angles. Deferoxamine B samples are shown as blue circles with areas proportional to the deferoxamine B concentration. (A) Mg/Si and Fe/Si ratios calculated for individual time points. (B) Relationship between the calculated Mg and Si release rates (μMols g −1 hr −1 ). The line corresponds to linear fit to both the control and deferoxamine B experiment. 2.4.3 Linearity of solute release with time Grouping all of the control experiments (up to∼340 hour duration) together and simultaneously fitting them with equation 2.2 shows that a linear model does not fully describe the relationship between solute release and time (adjusted R 2 =0.89; Figure 2.2a,c). Instead, the power-law model (equation 2.3) is a better match to the measurements (adjusted R 2 =0.98). While there are insufficient data to rigorously compare the power law and linear models for the deferoxamine B experiments, the linear model is a good descriptor of all the available data (all adjusted R 2 ≥ 0.96; Figure 2.2b,d, Table 2.1). 2.4.4 Dissolution rates For the linear model (equation 2.2), the dissolution rate is simply the calculated k linear value normalized by the stoichiometry of the element in the mineral phase. In the main text, I report only the average rate and propagated uncertainty (95% CI) for each experiment calculated by averaging the individual rates calculated for 15 0 100 200 300 0 2 4 6 8 10 Time (hours) Mg released ( μMols gram −1 ) 0 10 20 30 40 50 0 5 10 15 20 25 Time (hours) Mg released ( μMols gram −1 ) 0 100 200 300 0 2 4 6 8 Time (hours) Si released ( μMols g −1 ) 0 10 20 30 40 50 0 5 10 15 Time (hours) Si released ( μMols g −1 ) Mg = k power x time b + c adjusted R 2 = 0.98 Mg = k linear x time + a adjusted R 2 = 0.89 Mg = k power x time b + c adjusted R 2 = 0.98 Control Experiments Control Experiments Deferoxamine (point size = concentration) Control Experiments A B D C Deferoxamine (point size = concentration) Control Experiments Mg = k linear x time + a adjusted R 2 = 0.91 Figure 2.2: Solute release with time. Control samples are shown as trian- gles. Deferoxamine B samples are shown as circles with areas proportional to the deferoxamine B concentration. (A) Mg release with time for the∼340 hour con- trol experiments fit with both linear and power-law rate models (B) Comparison between Mg release in the control and deferoxamine B experiments. (C) Si release with time for the∼340 hour control experiments fit with both linear and power-law rate models (D) Comparison between Si release in the control and deferoxamine B experiments. Mg and Si. In Table 2.1 the rates calculated for each individual solute are reported. Typically, the individual rates overlap within their respective uncertainties. Combining the results of the control experiments and fitting only the data collected up until 40 hours with equation 2.2 yields a calculated rate of 0.03± 0.01 μmols olivine g −1 hour −1 . As the initial deferoxamine concentration is increased, rates calculated with equation 2.2 increase up to 0.29± 0.1 μmols olivine g −1 hour −1 for the maximum deferoxamine concentration of 281 μM (Figure 2.3). 16 0 100 200 300 0 0.02 0.04 0.06 0.08 0.1 Initial Deferoxamine ( μM) Fe release rate ( μMols g −1 hr −1 ) 0 100 200 300 0 0.2 0.4 0.6 Initial Deferoxamine ( μM) Mg release rate ( μMols g −1 hr −1 ) Control Experiments Deferoxamine A B Figure 2.3: Effect of deferoxamine B on olivine dissolution rates. Control samples are shown as triangles. Deferoxamine B samples are shown as circles. (A) Mg release rates (μMols g −1 hr −1 ) calculated using the linear rate model. (B) Fe release rates (μMols g −1 hr −1 ) calculated using the linear rate model. Combining the results of the control experiments and fitting all of the data (up until∼340 hours) with equation 2.3 yields rates that decreases with time following a power law function (Figure 2.2a). The power-law exponent calculated from equation 2.3 is 0.3± 0.1. 2.4.5 Spectrophotometric measurements Consistent with previous measurements (Farkas et al., 2001), the measured absorption spectrum for the Fe 3+ -deferoxamine complex has peaks at approxi- mately 250 and 430 nm (Figure 2.4). The addition of either Mg or Si to deferoxam- ine resulted in little change in the absorption spectrum relative to the deferoxam- ine only control (Figure 2.4). After reaction with olivine, the measured absorption spectrum contained peaks at 250 and 420 nm (Figure 2.4). 17 200 300 400 500 600 Fe 3+ - DFOB Mg 2+ / DFOB DFOB Si 4+ / DFOB Fe 3+ - DFOB DFOB reacted olivine 200 300 400 500 600 0.2 0.1 0.0 0.3 0.2 0.1 0.0 0.3 Wavelength (nm) Wavelength (nm) absorbance absorbance Figure 2.4: Ligand-metal complex absorption measurements. (A) Absorp- tion spectra for deferoxamine B in TRIS/HCl buffer (black) and with excess Si (blue), Mg (green), and Fe 3+ (red). (B) Comparison between the absorption spec- tra for the Fe 3+ -deferoxamine B complex (red) and a solution sample taken after reaction between deferoxamine B and olivine (green). 2.5 Discussion 2.5.1 Non-linear solute release and the mechanism of deferoxamine-promoted olivine dissolution Forthecontrolexperiments(i.e. thosewithoutaddedligands), dissolutionrates apparently decrease with time as indicated by the non-linear relationship between the masses of Mg and Si released and time (Figure 2.2a). This is unexpected since most previous experimental work has shown that at a fixed pH value, olivine dissolves at a constant rate that is independent of the aqueous activity of its constituentions(Pokrovsky&Schott,2000;Oelkers,2001;Olsen&Rimstidt,2008; Rimstidt et al., 2012). Many of these previous studies were conducted at a lower pH than in this study and with N 2 sparged fluids (i.e. lower pO 2 ). One batch 18 experiment by Chen & Brantley (2000), which was conducted under relatively similar conditions (pH 5 and and atmospheric pO 2 ), also showed a non-linear relationship between solute release and time, which was attributed to the effects of Fe-oxidation at the mineral surface. Intriguingly, the addition of deferoxamine in my experiments increases both the linearity and slope of the relationship between the mass of solute released and time relative to the control experiments (Figure 2.2). Based on these observations, I hypothesize that the non-linear behavior of the control experiments and its alleviation through deferoxamine addition is due to an inhibitory effect of Fe-oxyhydroxide precipitates that accumulate at the mineral surface when olivine dissolves under oxidizing conditions. This hypothesis is explored in more detail below. In all of the control experiments, Fe was not released stoichiometrically and was instead below the analytical detection limit of 0.5 μM (Figure 2.1), which implies that Fe released from the olivine structure rapidly precipitated as an oxyhydroxide mineral. As noted by Velbel (1993), a complete surface coating of Fe-oxides is unexpected during olivine dissolution since the molar volumes of typical Fe-oxide minerals are much lower than the molar volumes of forsterite (Mg end-member olivine) and fayalite (Fe end-member olivine). Instead of a complete surface coat- ing, I envisage a patchy surface coating that covers portions of the mineral surface. These shielded portions of the mineral surface may be less accessible to the bulk solution and, as a result, dissolve more slowly yielding a distribution of reaction rates, which is shown schematically in Figure 2.5a. While previous studies have yielded equivocal results as to whether or not surface coatings of Fe-oxyhydroxides can slow bulk mineral dissolution rates (Hodson, 2003; Ganor et al., 2005), they have only considered feldspar minerals, which do not contain significant structural Fe. The development of a patchy Fe-oxide surface coating can be modeled to see if it results in a temporal evolution in dissolution rates. For this model, I assume that olivine is composed of layers of Mg and Fe sites with each layer containing a fixed proportion of Fe sites that are randomly distributed across the layer. When a Mg siteisexposedtothesolution, itisimmediatelyreleased. WhenaFesiteisexposed to the solution, it is oxidized and leaves a surface Fe coating that can prevent the site below it from dissolving. I assume that the probability that the surface Fe coating inhibits dissolution is dependent upon the height of the surface coating 19 relative to the immediately adjacent sites and whether or not it is connected to other surface Fe coatings. Surface Fe coatings that are topographically high and unconnected have a low probability of inhibiting dissolution (Figure 2.5a). Surface Fe coatings that are topographically low and connected have a high probability of inhibiting dissolution (Figure 2.5a). In the model formulation, I identify 8 different classifications of surface Fe coatings depending upon their height and connectivity. Three of the classifications are assumed to have 100% probability of inhibiting dissolution for all model runs. The remaining classifications have variable probabilities that are proportional to the square of the sum of their height and connectivity. This model borrows concepts from the discussions presented in Gautier et al. (2001), Fischer et al. (2012), and Reeves & Rothman (2013) in that I consider a distribution of reaction rates. Further details of the model are presented in Appendix A. The model calculations show an initial power-law decay in the bulk dissolution rates with time due to the development of surface Fe 3+ coatings (Figure 2.5b). After some cutoff timescale, the bulk dissolution rates in the model simulations reach a constant value (Figure 2.5b). Both the exponent of the initial power-law decay and the cutoff timescale are dependent upon the model parameters that control the probability distribution of individual site dissolution rates (i.e. the number of Fe sites per surface layer and the probability that an Fe site with a given heightandconnectivitywillinhibitdissolution; Figure2.5b). Themodelprediction ofapower-lawdecayindissolutionratesisconsistentwiththeexperimentalresults, which can be described with a power-law rate model (adjusted R 2 = 0.98; Figure 2.2a,c). If olivine dissolution is inhibited by a patchy surface coating of Fe 3+ oxyhy- droxides at circum-neutral pH and atmospheric pO 2 , the presence of deferoxamine B in solution would be expected to remove this inhibition by dissolving the sur- face coating and stabilizing Fe 3+ as a dissolved ligand complex. This is consistent with the observed effect of deferoxamine B on Mg, Si, and Fe release rates (Figure 2.3) as well as the spectrophotometric measurements that show that a dissolved Fe 3+ -deferoxamine B complex is formed during the reaction (Figure 2.4). Exper- iments by Cheah et al. (2003) showed that the rate of deferoxamine-promoted Fe 3+ oxyhydroxide dissolution is proportional to the amount of surface adsorbed deferoxamine, which increases with increasing total deferoxamine concentrations 20 surface 0 20 40 60 80 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Time Step Relative Rate 2 4 −1.0 −0.5 0.0 log(Time Step) log(Relative Rate) y1 y2 y3 y4 x1 x2 x3 x4 x5 x6 Topographically high lower probability of inhibiting dissolution Connected higher probability of inhibiting dissolution timestep 1 timestep 2 timestep 3 timestep 4 solution Fe 2+ site Fe 3+ site Mg 2+ site A B Lower probability of inhibition Higher probability of inhibition C Figure 2.5: Model of the effects of surface coatings on dissolution rates. (A) Schematic showing the model design and hypothesized effects of different sur- face coatings on dissolution rates. (B) Model results for a simulation with a higher probability of inhibition (larger range in T; gray points) and lower probability of inhibition (smaller range in T; black points). The points and error bars correspond to the mean values and standard deviations calculated for five separate model runs. (C) log-log plot of the data shown in panel B. The regression lines highlight the initial power-law decay in rates. following a Langmuir adsorption relationship. These observations are broadly con- sistent with the data, which show a non-linear relationship between dissolution rates and initial deferoxamine concentrations (Figure 2.3). As mentioned above, most previous olivine dissolution experiments, including some that tested the effects of deferoxamine B (e.g., Wolff-Boenisch et al. 2011), were performed under conditions that would be expected to slow or prevent the oxidation of Fe 2+ released from olivine (i.e. low pH and/or low pO 2 ). Both the thermodynamics and kinetics of Fe 2+ oxidation are extremely sensitive to pH and pO 2 with lower values of both these parameters favoring the stabilization of dis- solved Fe. Potentially, this explains why non-linear olivine dissolution rates are not typically reported. However, experiments by Sugimori et al. (2012), which specifi- cally focused on the whether or not Fe 2+ oxidation affects olivine dissolution rates, 21 suggest that Fe 3+ oxide precipitates do not affect bulk dissolution rates. Poten- tially, the discrepancy between the results of this study and those of Sugimori et al. (2012) is due to the different pH buffers utilized. In their study, Sugimori et al. (2012) utilized a 40 mM acetate buffer, which is known to accelerate Fe release from fayalite (Daval et al., 2010). Alternatively, there may be a sharp pH thresh- old such that the experiments by Sugimori et al. (2012), which were performed at a maximum pH of 6, would not be expected to be affected by Fe-oxyhydroxide precipitation. I think this is unlikely since the results of an experiment at pH 5.5 by Chen & Brantley (2000) are consistent with my results/hypothesis. 2.5.2 Implications for microbially-accelerated silicate min- eral dissolution The experimental results of this work and previous studies (e.g., Page & Huyer 1984; Liermann et al. 2000; Haack et al. 2008; Shirvani & Nourbakhsh 2010) sug- gest that the dissolution rates of a wide range of Fe-silicates with different chem- ical compositions, mineral structures, and dissolution mechanisms all increase in response to increasing siderophore concentrations. The available data show that rates are significantly accelerated at micromolar siderophore concentrations (Fig- ure 2.3, Liermann et al. 2000; Haack et al. 2008; Shirvani & Nourbakhsh 2010). In contrast, environmental concentrations are typically orders of magnitude lower (nM range; Powell et al. 1980; Holmström & Lundström 2004; Essén et al. 2006; Vraspir&Butler2009). Experimentalresultssuggestthattherelationshipbetween dissolutionrateaccelerationandsiderophoreconcentrationisrelatedtotheadsorp- tion isotherm of the siderophore to the mineral surface (Cheah et al., 2003). Organic surfactants, which can be produced by microorganisms, have been shown to increase the sensitivity of goethite (α-FeOOH) dissolution to siderophore con- centrations by affecting the partitioning between dissolved and adsorbed species (Carrasco et al., 2007), but the magnitude of this effect is less than the difference between environmental siderophore concentrations and the siderophore concentra- tions required for maximum dissolution rates. While the simple discussion presented above suggests that siderophores may not play an important role in the microbial acceleration of natural silicate mineral dissolution, it assumes that the measurements of bulk siderophore concentrations 22 (e.g., in soil solutions) are equivalent to the concentrations at mineral surfaces. It is conceivable that siderophores are concentrated within microbial biofilms formed around mineral surfaces. In this case, the bulk siderophore concentration measure- ments may underestimate the effects of siderophores on dissolution rates. However, even withinthe pore fluid of a biofilm, there exists alimit on maximum siderophore concentrations due to the feedbacks between Fe concentrations, siderophore pro- duction, and microbial growth. Since siderophore production decreases in response to increasing Fe concentrations (Fekete et al., 1983; Sweet & Douglas, 1991; Wil- helm et al., 1996; Berraho et al., 1997), there should be a negative feedback that regulates the extent to which the dissolution of natural silicate minerals is accel- erated via siderophore production. To explore the feedbacks between Fe concentrations, siderophore production, and microbial growth and how they might regulate the extent of siderophore- promoted dissolution, I first consider simple mathematical models of microbial growth and siderophore production. To a first order, growth rates (μ) can be modeled to increase as the concentration of a limiting nutrient (e.g., Fe) increases following the Monod equation: μ =μ max × C s k s +C s (2.6) where μ max is the maximum growth rate, C s is the concentration of the limiting nutrient, and k s is a constant equal to the concentration of the limiting nutrient at which rates are half of their maximum value. This model assumes that the relationship between the mass of Fe required for a unit of biomass is constant. Experimental data of specific siderophore yield (μmol siderophore/gram biomass; Y sid ) show a decrease with increasing Fe concentrations (Fekete et al., 1983; Sweet & Douglas, 1991; Berraho et al., 1997). A convenient mathematical model of this relationship is: Y sid = −Y max × [Fe] k L + [Fe] +Y max (2.7) where Y max is the maximum specific siderophore yield and k L is a constant equal to the concentration at which Y sid is half of its maximum value. For example, the data of Fekete et al. (1983) can be modeled with equation 2.7 (adjusted R 2 = 0.98; 23 Figure 2.6). Together, the product of equations 2.6 and 2.7 is the total siderophore production rate. Since μ is positively correlated with Fe concentrations and Y sid is negatively correlated with Fe concentrations, total siderophore production rates show a ’humped’ relationship where there is a peak in the rate of siderophore pro- duction at intermediate Fe concentrations (Figure 2.6). While the exact shape of the relationship between total siderophore production rates and Fe concentrations will be sensitive to the mathematical models of microbial growth and siderophore yield utilized, I assert that the general notion of a ’humped’ relationship is theo- retically sound. At extremely low Fe concentrations, microbial growth should be limited to such an extent that total siderophore production rates will be negligible. At very high Fe concentrations, siderophore production is unnecessary and should cease. I will continue the analysis using equations 2.6 and 2.7 since they appear to be reasonable empirical descriptors of the available data (Figure 2.6; Sweet & Douglas 1991; Berraho et al. 1997). 5 10 15 20 25 2 4 6 8 10 12 14 16 0 5 10 15 20 25 6 8 10 12 14 16 30 2 0 0.5 1 1.5 2.5 Fe (μM) mg biomass hr -1 μMols siderphore mg -1 Fe (μM) 0 μMols siderophore hr -1 A. vinelandii (Fekete et al. 1983) A B Figure 2.6: Fe-dependent microbial growth and ligand production. (A) Data from Fekete et al. (1983) showing increasing growth rates with increasing Fe concentrations (blue) and decreasing siderophore yield (μMols siderophore mg −1 ) within increasing Fe concentrations (green). Regression lines correspond to equa- tions 2.6 and 2.7. (B) Siderophore production rates as a function of Fe concentra- tions calculated based on the regression lines shown in panel A. I hypothesize that the shape of the relationship between total siderophore production rates and Fe concentrations is an important control on microbially- accelerated silicate mineral dissolution. If the peak in total siderophore production rate for a hypothetical microorganism is at relatively high Fe concentrations, then 24 this should allow silicate mineral dissolution rates to reach higher values before siderophore production slows due to an overabundance of Fe. If the peak in total siderophore production rate for a hypothetical microorganism is at relatively low Fe concentrations, then dissolution rates will only increase marginally before total siderophore production rates slow. 2.5.3 Feedback Model The hypothesis presented above can be tested by combining equations 2.6 and 2.7 with a model of siderophore-promoted Fe release from solid phases. For this model, which is described in detail in Appendix B, I treat the system as a well- mixed flow-through reactor. While natural systems are often not "well-mixed", I formulate the model in this way to specifically focus on the feedbacks between microbial physiology and mineral dissolution kinetics. I acknowledge that the cou- pling between reaction and transport processes can also influence the effects of organic ligands on mineral dissolution processes (e.g., Lawrence et al. 2014). Simi- larly, I assume that the solution pH remains constant despite changes in microbial growthratesandsilicatemineraldissolutionrates. Whileenhancedsilicatemineral dissolution should lead to increases in pH, the effects of microbial metabolisms on pH are much more varied (e.g., Soetaert et al. 2007) and can depend strongly on the chemical composition the solution (Meister, 2013). Consequently, the model I present is not meant to fully reproduce weathering processes in natural systems, but is used instead to explore how the feedback between the production of lig- ands and the release of a limiting nutrient controls the extent to which mineral dissolution rates are accelerated. The results of a non-dimensional version of the model (see Appendix B) show that both microbial physiology and the timescale over which chemical reactions occur influence the extent to which reaction rates are accelerated via siderophore production (Figure 2.7). In Figure 2.7, the reaction timescale is represented by the non-dimensional parameter A (see Appendix B) where A = μ×Y max k s ×Q (2.8) 25 k s is the ligand concentration at which dissolution rates are half of their maximum values, and Q is the flow rate in and out of the system. This parameter is a Damköhler number (i.e. the ratio of a chemical reaction rate to the physical transport rate) calculated for the theoretical maximum siderophore production rate. Low A values reflect conditions where physical transport is fast relative to siderophore production, which limits the accumulation of siderophores in solution and the extent to which reaction rates are accelerated (Figure 2.7). High A values reflect physical transport rates that are much slower than reaction rates. At highA values, there is ample time available for reaction between siderophores and mineral phases, which also limits the extent to which reaction rates are accelerated (Figure 2.7). At intermediate A values, there is a peak in extent to which siderophore production accelerates reaction rates that occurs because physical transport is fast enough to enough to keep Fe concentration low, but does not limit siderophore production (Figure 2.7). 10 0 10 1 10 2 0 0.05 0.1 0.15 0.2 0.25 ( μ max Y max ) / (Q × k S ) % of max reaction rate 10 0 10 1 10 2 0 0.05 0.1 0.15 0.2 0.25 ( μ max Y max ) / (Q × k S ) S* Ref kL = 0.5 x Ref kL = 1.5 x Ref Y max = 0.5 x Ref Y max = 1.5 x Ref A B Figure 2.7: Siderophore production feedback model. (A) Calculated steady- state non-dimensional siderophore concentrations (S*) for models with different Y max and kL values relative to the reference values, which are based on the data of Fekete et al. (1983) (Figure 2.6). (B) The percentage of the maximum dissolution rate the corresponds with the calculated S* value. See section B for a detailed explanation of the model calculations For the range of parameter values explored, the maximum steady-state non- dimensional siderophore concentrations (S*) correspond to mineral dissolution rates reaching 10-20% of their maximum values (Figure 2.7b). The maximum S* values vary between simulations with different Y max and k L values. Higher values, 26 which correspond to higher siderophore yields (Y max ) at higher Fe concentration (k L ), increase S* values, confirming the importance of microbial physiology on the extent to which dissolution rates are accelerated (Figure 2.7). 2.6 Conclusions The results of this study show that at circum-neutral pH, olivine dissolution rates are appreciably accelerated by the model siderophore deferoxamine B (Figure 2.3). Based on the observed decrease in dissolution rates with time for the ligand- free control experiments but not for the experiments with siderophores (Figure 2.2), I hypothesize that siderophores affect olivine dissolution rates by preventing the formation of a surface layer of Fe-oxyhydroxides that shield portions of the mineral surface from contact with the solution. This hypothesis is consistent with the results of a simple mineral dissolution model (Figure 2.5) as well as previous research indicating that the formation of a surface coating of Fe 3+ -silicate minerals inhibits the dissolution of fayalite (Santelli et al., 2001). Extrapolating the results of this and other studies to natural systems requires careful consideration of how nutrient concentrations affect biological growth and ligand production (Figure 2.6). The results of a simple feedback model show that theextenttowhichFe-limitationcanacceleratereactionsratesdependsonboththe timescale over which reactions occur as well as physiological parameters that affect the amount of siderophores produced as a function of Fe concentrations (Figure 2.7). Furtherresearchintothecontrolsonsiderophoreproductioninenvironmental microorganisms is likely to prove useful to determining the effects of microbial activity on mineral dissolution rates. Deferoxamine B Mg rate ± R 2 Si rate ± R 2 Fe rate ± R 2 21.59 0.25 0.19 0.99 0.16 0.15 0.99 0.04 0.04 0.99 104.98 0.35 0.00 1.00 0.23 0.26 0.98 0.06 0.04 0.99 214.97 0.54 0.12 1.00 0.32 0.01 1.00 0.08 0.02 1.00 141.56 0.49 0.06 1.00 0.26 0.03 1.00 0.06 0.01 0.99 26.66 0.18 0.09 0.96 0.10 0.05 0.96 0.02 0.02 0.91 281.16 0.52 0.14 0.99 0.29 0.07 0.99 0.06 0.02 0.99 0.00 0.06 0.02 0.90 0.03 0.01 0.86 0.00 0.00 0.00 Table 2.1: Calculated Linear Dissolution Rates 27 Chapter 3 Geomorphic regime modulates hydrologic control of chemical weathering in the Andes-Amazon 3.1 Linking Statement This chapter reports the results of a field study on how chemical weathering is affected by the transport of water and minerals at Earth’s surface. As alluded to in section 2.1, there exists a major discrepancy between mineral reaction rates measured in field and laboratory settings. While no single mechanism appears to completely account for this offset, some of the burden may be placed on the fact that we do not fully understand the timescale over which water reacts with minerals in surface environments (i.e. fluid transit times; Kirchner 2015) and how minerals "age" with continued reaction (White & Brantley, 2003; Keech et al., 2013). The design of the field study presented in this chapter is well suited to explore both of these issues in natural systems since it examines weathering over a rangeofhydrologicconditionsalongagradientofweatheringduration(i.e. mineral residence times). While we still lack exact tools to fully quantify fluid transit times and mineral residence times in natural systems, this chapter utilizes a novel proxy, the concentration-runoff relationship, which records the effects of both timescales and is directly linked to chemical weathering fluxes (Godsey et al., 2009; Maher, 2011; Maher & Chamberlain, 2014). Dr. Kathryn Clark is acknowledged for not only collecting all of the samples, but also generating all of the hydrologic data with some additional assistance from Natalie Ballew. Again, Dr. Joshua West provided key insights and supervision. A manuscript based on this work was accepted for publication in Geochimica et Cosmochimica Acta after a thorough review by Dr. Julien Bouchez and two anony- mous reviewers. 28 Financial support for this research was provided by NSF EAR-1227192 to A.J. West. K. Clark was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC) and Clarendon Fund PhD scholarships. M. Torres was supported by USC College doctoral and C-DEBI fellowships. I thank ACCA Peru(Wayqecha,CICRA),Incaterra(SanPedro),andCREES(MLC)forfieldsup- port; Arturo Robles Caceres, J.A. Gibaja Lopez, J. Huamán Ovalle, R.J. Abarca Martínez, I. Cuba Torres, A. Alfaro-Tapia, D. Oviedo Licona, and J. Farfan Flores for field assistance; and Julien Bouchez, Jotautas Baronas, and James Kirchner for helpful discussions prior to submission. 3.2 Introduction The chemical transformation of minerals in Earth’s critical zone (Brantley et al., 2014) plays an important role in many processes such as soil development (Graham et al., 2010), cycling of essential nutrient elements (Chadwick et al., 1999; Porder et al., 2007), and regulation of Earth’s climate via the transfer of carbon between the surficial and rock reservoirs over geologic timescales (Walker et al., 1981; Berner et al., 1983; Beaulieu et al., 2012). It has long been recognized that the supply, storage, and routing of water influences the rate of dissolution of primary minerals and precipitation of secondary phases. The importance of hydrology is readily apparent from the observation that concentrations of dissolved ions derived from weathering are correlated with runoff and rarely show simple dilution with increasing flow in streams and rivers (Godsey et al., 2009; Moon et al., 2014). Consequently, chemical weathering fluxes are closely tied to changes in precipitation and/or river runoff (Dunne, 1978; Bluth & Kump, 1994; Anderson et al., 1997; Gaillardet et al., 1999; West et al., 2005; Gabet et al., 2006; Hartmann et al., 2009; Maher, 2011; Eiriksdottir et al., 2013). The strength of the correlation between solute concentration and runoff (i.e. the "concentration-runoff relationship") varies substantially between different ele- ments (Kirchner & Neal, 2013; Moon et al., 2014). In many catchments, the concentrations of base cations (i.e. Ca, Mg, Na, and K) and silicon are highly correlated with runoff and exhibit a "chemostatic" response to flow where the vari- ation in solute concentrations is substantially smaller than the variation in runoff 29 (Godsey et al., 2009; Clow & Mast, 2010; Stallard & Murphy, 2014). However, concentration-runoff relationships for base cations and silicon vary from one river to another, ranging from effectively chemostatic behavior to significant dilution with increasing runoff (Moon et al., 2014). While multiple mechanisms can generate the concentration-runoff relationships for weathering-derived solutes observed in natural systems (Johnson et al., 1969; Christophersen et al., 1990; Godsey et al., 2009; Maher, 2011; Maher & Cham- berlain, 2014), all of the mechanisms share some basic principles. Fundamentally, runoff rates reflect the supply, storage, and routing of water in the catchment sys- tem (Beven, 2011). These factors affect the amount of time it takes precipitation to transit through the catchment and become river flow (i.e. the fluid transit time) as well as the physical and chemical properties of the flow path that fluids take on their way to the channel (e.g., the water saturation, porosity, and primary min- eral abundances). The fluid transit time and the flow path properties can affect chemical weathering reactions by modifying the time over which the reaction can occur and the chemical and physical properties of the material undergoing reac- tion respectively. Consequently, concentration-runoff relationships should, to a first order, reflect the hydrologic controls on chemical weathering fluxes in catch- ment systems (Godsey et al., 2009; Maher, 2011; Maher & Chamberlain, 2014). Multiple mechanisms can generate runoff in a single catchment system depend- ing upon precipitation fluxes and antecedent conditions (Beven, 2011). As a result, concentration-runoff relationships may show temporal variability within a single catchment. For example, runoff generation during large storms may dominantly result from flow through near-surface flow paths where fluids interact minimally with primary reactive minerals (Bonell & Gilmour, 1978; Calmels et al., 2011; Stallard & Murphy, 2014). In contrast, river water during base flow conditions may dominantly be sourced from deeper subsurface flow paths with longer transit times (Beven, 2011; Calmels et al., 2011). This systematic change in the dominant runoff generating mechanism at high runoff could considerably affect weathering fluxes because storm events, which transport the large volumes of water, may not transit through the zone of active weathering and acquire solutes. To identify the importance of such processes, measurements of concentration and runoff need to be made over the entire range of hydrologic conditions. 30 Ratios of solute concentrations of different elements are a particularly useful tool for investigating the underlying processes responsible for supplying solutes to rivers. The weathering of different lithologies produces distinct solute elemental ratios that can be used to determine the relative contributions from different litho- logic sources (e.g., Sr/Na and Ca/Na ratios; Negrel et al. 1993; Gaillardet et al. 1999). Secondary processes, such as the removal of solutes via the precipitation of minerals or uptake by organisms, discriminate between elements and lead to characteristic variations in solute elemental ratios (e.g., element/Si ratios; Derry et al. 2005; Georg et al. 2007; Kurtz et al. 2011. Because weathering reactions produce depth-dependent variations in the chemistry and structure of the criti- cal zone (e.g., Lebedeva et al. 2007), flow paths that sample different depths can produce fluids with unique solute elemental ratios (Boy et al., 2008; Kurtz et al., 2011). By investigating how elemental ratios vary with runoff (i.e. "ratio-runoff relationships"), hydrologically driven changes in biogeochemical processes and flow paths can be identified (Tipper et al., 2006; Boy et al., 2008; Calmels et al., 2011; Kurtz et al., 2011). In this work, I consider how concentration-runoff and ratio-runoff relationships varyacrossageomorphicgradient, andIexplorewhatthisvariabilityimpliesabout the links between chemical weathering, hydrology, and erosion. I focus on the transition from the steep slopes and rapid erosion in the Andes to the depositional settingoftheAmazonforelandfloodplainofPeru. Variationingeomorphicregime, such as across the study area, provides a natural gradient in the timescale over which minerals are exposed to chemical weathering reactions. In rapidly eroding environments, minerals undergo chemical reactions for relatively short periods of time before being exported from the catchment (i.e. short mineral residence time; West et al. 2005; Gabet & Mudd 2009; Ferrier & Kirchner 2008; Hilley et al. 2010; West 2012). In floodplain environments, sediment storage processes produce distributions of mineral residence times with relatively long means and heavy tails (Bradley & Tucker, 2013). Variation in mineral residence times should lead to variation in the abundance of primary reactive minerals and, consequently, the rate at which solutes can be generated within the critical zone (West, 2012; Maher & Chamberlain, 2014). 31 While both geomorphic and hydrologic processes are thought to modulate chemical weathering fluxes, the question remains how these processes may inter- act in natural systems (Maher & Chamberlain, 2014; Li et al., 2014). If fluid transit times and mineral residence times co-vary in natural systems, variations in chemical weathering fluxes may be amplified or dampened between different envi- ronments. Potentially, the development of topography may act to increase fluid transit times by creating longer fluid flow paths (e.g. McGuire et al. 2005). If fluid transit times and topography are correlated this way in other systems, then mountainous regions may be efficient chemical weathering systems because they have relatively long fluid transit times, short mineral residence times, and large water fluxes driven by orographic precipitation (Gaillardet et al., 2011; Li et al., 2014; Maher & Chamberlain, 2014). This study takes advantage of samples taken at high temporal resolution (monthly to sub-daily sampling) across a relatively small area (27830 km 2 or 0.6% of the entire Amazon basin) in the Andes and Amazon. The study design makes it possible to limit heterogeneity in lithology (which is known to exert a strong control on weathering processes; e.g., Bluth & Kump 1994; Gaillardet et al. 1999) while still considering very significant contrasts in erosional setting. I report time-series data of dissolved ion concentrations and use these data to determine how concentration-runoff relationships and ratio-runoff relationships of weathering-derived solutes vary between catchments. These relationships are used to infer how the hydrologic control on chemical weathering responds to variation in the geomorphic regime. 3.3 Methods 3.3.1 Study Site The sample set used in this study comes from the Kosñipata Valley and Madre de Dios floodplain in Peru (Figure 3.1). Four nested catchments spanning over 4000 meters in elevation and 120 km in horizontal distance were sampled over the course of one year in order to capture the transition from Andes Mountains (> 400 m elevation; cf. Moquet et al. 2011) to the foreland floodplain (120 to 400 m elevation; Moquet et al. 2011). A consequence of the nested catchment 32 −72˚ −70˚ −12˚ 0 1000 2000 3000 4000 5000 Elevation (m) 100 km Way 3204 m 50 km 2 SP 2778 m 160 km 2 MLC 2012 m 6025 km 2 CICRA 822 m 27830 km 2 −90˚ −60˚ −30˚ -30˚ 0˚ A. B. Figure 3.1: Map of the study site. (A) The points indicate the approximate locations of the time-series sampling localities studied along the Kosñipata-Madre de Dios river system in the Peruvian Andes and lowland Amazon. The areas upstream of each of the four locations define the four nested catchments, with mean elevation and total area for each of the catchments reported. (B) Inset showing the location of the study region within the larger Amazon system. design is that both mean catchment elevation and total catchment surface area are negatively correlated. The catchment properties presented below and in Table 3.1 (i.e. mean elevation, elevation range, and mean catchment slope angle) were all calculated using a digital elevation model derived from Shuttle Radar Tomography Mission data (90 x 90 m resolution; Jarvis et al. 2008). The mean catchment slope angles for the Andean sites (Way and SP; see below) calculated in this study are 1 ◦ lower than those calculated in Clark et al. (2014) for the same catchments, which is likely due to different post-processing of the SRTM data. Because Clark et al. (2014) did not calculate mean catchment slope angles for the lower elevation sites (MLC and CICRA; see below) I only use the mean catchment slope angles calculated in this study to be consistent when comparing between sites. 33 Andean Sites To constrain weathering processes occurring in the Andes, the Kosñipata River was sampled at the Wayqecha gauging station (Way) and at the San Pedro gauging station(SP;Figure3.1). ThecatchmentupstreamoftheWayqechagaugingstation (S13 09.732 W71 35.333) has a total area of 50 km 2 , an elevation range of 2250- 3910 m, a mean elevation of 3204 m, and a mean catchment slope angle of 25 ◦ (Figure 3.1; Clark et al. 2013). The catchment upstream of the San Pedro gauging station (S13 03.444 W71 32.664) has a total area of 161 km 2 , an elevation range of 1360-4000 m, a mean elevation of 2778 m, and mean catchment slope angle of 27 ◦ (Figure 3.1; Clark et al. 2013). The dominant lithology in both the Way and SP catchments is a Paleozoic (∼450 Ma) metasedimentary mudstone unit with some carbonate cements (Car- lottoCaillauxetal.,1996). Aminorproportion(i.e. 21%byarea)ofthecatchment defined at San Pedro gauging station is underlain by Paleozoic felsic plutonic rocks (Carlotto Caillaux et al., 1996; Clark et al., 2013). Soils in Andean landscapes have been previously classified as Inceptisols (Asner et al., 2014). The mean annual air temperature in the Andean catchments ranges from∼12 to 19 ◦ C, varying with the elevation (Girardin et al., 2013; Huasco et al., 2014), with an adiabatic air temperature lapse rate of 4.94 ◦ C km −1 (Girardin et al., 2013). Catchment-wide mean annual rainfall values of 2299± 115 mm yr −1 (Way) and 2881± 124 mm yr −1 (SP) were determined for the period from 1998 to 2012 using calibrated TRMM data (Clark et al., 2014). During the study period (2010- 2011), annual rainfall was 2519± 335 mm yr −1 and 3112± 414 mm yr −1 for the Way and SP catchments respectively, slightly above the long-term average (Clark et al., 2014). Like air temperature, rainfall also shows a strong elevational dependence(Lambsetal.,2012;Clarketal.,2014). Forthesame2010-2011period, annual runoff was estimated to be 3065 mm yr −1 and 2796± 126 mm yr −1 for the Way and SP catchments respectively. The discrepancy between rainfall and runoff at the Way catchment reflects, in part, the contribution of cloud water inputs to the total water budget and the fact that the annual discharge at this site is highly uncertain as it was calculated using 59 discrete measurements as opposed to the near-continuous discharge logging that was conducted at San Pedro (i.e. 4 measurements per hour). For both of the Andean catchments, a seasonal hysteresis 34 between rainfall amounts and runoff is evident with runoff during the dry season (May-September) being sustained by relatively lower amounts of rainfall. Likely, this hysteresis reflects seasonal storage of rainfall as groundwater within fractured bedrock (Clark et al., 2014). Mountain Front Site To constrain weathering processes occurring over the entire range of elevations within the Peruvian Andes that contribute to the Amazonian weathering budget, the Alto Madre de Dios River was sampled at the Manu Learning Center (MLC). The MLC catchment (S12 47.334 W71 23.429) has a total area of 6025 km 2 , an elevation range of 452-5496 m, a mean elevation of 2012 m, and mean catchment slope angle of 22 ◦ C (Figure 3.1). The catchment area upstream of MLC is largely underlain by Paleozoic metasedimentary rocks, Paleozoic plutonic rocks, and Pale- ozoic marine sedimentary rocks (Mendívil Echevarría & Dávila Manrique, 1994; Carlotto Caillaux et al., 1996; Vargas Vilchez & Hipolito Romero, 1998; INGEM- MET, 2013) Foreland Floodplain Site To constrain weathering processes in the foreland floodplain region, the Madre de Dios River was sampled at the CICRA research station (Figure 3.1). The CICRA catchment (S12 34.362 W70 06.049) has a total area of 27830 km 2 , a mean elevation of 822 m, elevation range of 219-5496 m, and a mean catchment slope angle of 9 ◦ (Figure 3.1). Because the CICRA catchment incorporates both the Andes and foreland-floodplain region, temperature is heterogeneous throughout the catchment. Near the sampling locality, meteorological stations (data available from http://atrium.andesamazon.org) record temperatures that typically range from 21-26 ◦ C with a mean of∼24 ◦ C and precipitation volumes that range from 2600 to 3500 mm yr −1 , with mean values between 2700 and 3000 mm yr −1 (Lambs et al. 2012). A majority of the annual rainfall in Peru occurs during the austral summer (Lambs et al., 2012; Clark et al., 2014). The catchment area of CICRA is largely underlain by sediments shed from the Andes with additional contributions from Cretaceous marine sediments and Cenozoic continental deposits (Räsänen et 35 al., 1992; Mendívil Echevarría & Dávila Manrique, 1994; Carlotto Caillaux et al., 1996; Vargas Vilchez & Hipolito Romero, 1998; INGEMMET, 2013) Gauge Station Wayqecha (Way) San Pedro (SP) Manu Learning Center (MLC) Los Amigos (CICRA) Latitude S13 09.732 S13 03.444 S12 47.334 S12 34.362 Longitude W71 35.333 W71 32.664 W71 23.429 W70 06.049 Area (km 2 ) 50 161 6025 27830 mean elevation (m) 3204 2778 2012 822 median elevation (m) 3242 2795 1830 445 min elevation (m) 2250 1360 452 219 max elevation (m) 3910 4000 5496 5496 mean slope angle (degrees) 25 27 22 9 min slope angle (degrees) 0.64 0.55 0 0 max slope angle (degrees) 76.97 76.97 54.5 49.07 number of samples 86 60 52 23 Table 3.1: Catchment Metadata 3.3.2 Sampling Methodology Samples were collected between 2010 and 2011 from each catchment at approx- imately bi-weekly resolution as well as every three hours for a period of about two weeks starting on January 30th in 2010. However, only data collected with paired discharge measurement are reported here for the main sampling localities from the 2010-2011 sampling campaign. A total of 86, 60, 52, and 23 paired discharge and dissolved element samples were collected from Way, SP, MLC, and CICRA respec- tively. Because the Wayqecha gauging station is immediately downstream of a confluence of two tributaries, the east (Alipachanca) and west (Huallpayuncha) 36 tributaries were also sampled periodically for chemical analysis. Both tributaries were sampled at the same time as the main stem downstream of the confluence (i.e. at the Wayqecha gauging station) in order to investigate mixing relationships. Following Clark et al. (2014), samples collected between December and March are labeled wet season samples, samples collected in April are labeled wet-to-dry sea- son transition samples, samples collected between May and September are labeled dry seasons samples, and samples collected between October and November are labeled dry-to-wet season transition samples. Samples for major and minor element analysis were collected from the river surface in a clean polypropylene (PP) bottle, filtered onsite with a 0.2 μm nylon filter and split into two aliquots stored in separate 60 mL high-density polyethylene bottles (HDPE). One aliquot was preserved with 2 drops of high purity HCl dis- pensed from an acid-washed Teflon dropper bottle for cation analyses. The other aliquot was left unpreserved for anion analysis. In the laboratory, samples with any remaining particulates (e.g., from flocculated aggregates forming after field filtration) were re-filtered before analysis with a 0.2 μm nylon filter. Discharge was measured at each of the main sampling localities as described by Ballew (2011) and Clark et al. (2014). Briefly, river level was monitored manually at all sites and converted to discharge using a rating curve. A total of 28, 22, 20, and 11 points were used for the rating curve at the Way, SP, MLC, and CICRA catchments respectively. For the San Pedro site, river level was also monitored with a water level logger that recorded river level measurements every 15 minutes. 3.3.3 Analytical Methodology Major and minor cation (Ca, Mg, Na, K, Si, Li, and Sr) concentrations were measured on the acidified aliquot using an Agilent Microwave Plasma-Optical Emission Spectrometer (MP-OES). Precision and accuracy was assessed by analyz- ingareferencematerialafterevery15samples. ForCa, Mg, Na, K,Si, thereference material ION-915 (Environment Canada) was used. For Li, the reference material TMDA-51.4 (Environment Canada) was used. For Sr, an in-house prepared SrCO 3 solution was used. Replicate analyses of each solution reveal analytical precision within 5% (1σ) for each analyte. 37 Major anion (SO 4 2− and Cl − ) concentrations were measured on the un-acidified aliquot with a Metrohm Ion Chromatograph equipped with a Metrosep A4/150 column and conductivity suppression. The elements were eluted with 3.2 mM Na 2 CO 3 and 1.0 mM NaHCO 3 at a flow rate of 0.7 mL min −1 . Precision and accuracy was assessed by analyzing a certified reference material (ION-915, Envi- ronment Canada) after every 15 samples. Replicate analyses of ION-915 reveal an analytical precision within 5% (1σ) for each analyte. 3.3.4 Data Processing Correction for cyclic salt inputs Atmospheric deposition has been shown to be an important source of major ele- ments in the Amazon system (e.g., Stallard & Edmond 1981; Moquet et al. 2011). Tocorrectforthiscontribution,precipitationelement/Cl − ratiosaretypicallyused. For this study, samples of local precipitation were collected and analyzed, and the new precipitation chemistry data was supplemented with previously published pre- cipitation analyses (Stallard & Edmond, 1981; Moquet et al., 2011). Because both elemental concentrations and ratios in precipitation show large variations (Figure 3.2), precipitation volume-weighted means are typically used to define the chemical composition of precipitation (e.g., Calmels et al. 2011). Because paired measure- ments of precipitation chemistry and volume for the studied catchments are not currently available, an alternative approach was utilized here. This method, which is detailed in the Appendix C, relies on the assumptions that 1) stream water reflects a two component mixture of solutes derived from weathering and precipi- tation and 2) seawater aerosols represent the dominant source of Na and Cl − ions to precipitation in the system (Stallard & Edmond, 1981; Andreae et al., 1990b). Calculation of concentration-runoff relationships Following Godsey et al. (2009), the relationship between concentration and runoff (Q) was modeled with a power law function where: Concentration =a× (Q) b (3.1) 38 0.1 1 10 100 Na/Cl Mg/Cl 0 10 20 30 40 50 60 0 10 20 30 40 50 60 Na/Cl Mg/Cl rain (this study) rain (literature) San Pedro river sample best fit rain Mg/Cl ratio and 95% CI 0.1 1 10 100 0.01 Figure 3.2: Example of the atmospheric input correction method. The end-member Mg/Cl − ratio of precipitation for San Pedro was calculated using a linear fit to the measured Na/Cl − and Mg/Cl − values in the river samples, and the assumed Na/Cl − precipitation end-member ratio of 0.857. (A) A plot of Na/Cl − versus Mg/Cl − for samples from the San Pedro catchment with logarithmically scaled axes. Both river water and rainwater samples are plotted along with the calculated best-fit value for the atmospheric input Mg/Cl ratio and its 95% con- fidence interval. (B) A plot of Na/Cl − versus Mg/Cl − with linearly scales axes showing the degree to which a linear function describes the riverwater data. The same approach was used to determine other element/Cl − ratios in precipitation for all of the studied catchments (i.e. in addition to Mg, shown here). Non-linear regression of the data using equation 3.1 and calculation of the uncer- tainties(i.e. 95%CI)associatedwiththefittedparameters(aandb)wasperformed using thetrust-regionalgorithm available inthe MATLAB 2013acurvefitting tool- box. An arbitrary degrees of freedom-adjusted R 2 value of 0.3 was set in order to identify solutes that were correlated with runoff. For all of the catchments, analy- sis of the residuals of the power law fit revealed that a single concentration-runoff relationship misestimated the concentrations of specific subset of the data. So, in additiontofittingthedatawithequation3.1usingallthedatapoints, fitswerealso calculated with subsets of the data that were selected based on criteria described in the results section. In all tables, the coefficients labeled "total fit" were calculated using all of the data points, the coefficients labeled "dominant fit" were calculated using a subset of the data pointed selected to reflect the dominant trend, and, for somesites, thecoefficientslabeled"subordinatefit"werecalculatedwithasubsetof 39 the data points that showed a subordinate concentration-runoff relationship. Each catchment ultimately required different criteria to separate data points described by different concentration-runoff relationships (see section 3.4.2 below). The cal- culated values of the power law exponent and their 95% confidence intervals are presented in Table 3.2. Both uncorrected and atmospheric-input corrected data were fit with equation 3.1 in order to calculate concentration-runoff relationships for the concentrations of elements sourced solely from weathering reactions (i.e. the atmospheric-input corrected data) and compare them to concentration-runoff relationships for total solute concentrations (i.e. the uncorrected data). Calculation of ratio-runoff relationships For each of the studied catchments, variation in both the atmospheric input corrected and uncorrected Sr/Na ratios with runoff were fit with a linear model using MATLAB 2013a. The uncertainties (i.e. 95% CI) of the calculated slopes were used to assess whether or not the elemental ratios varied significantly with runoff(i.e. non-zeroslopes). Alinearregressionwasalsousedtotestforcorrelation betweenNa/SiandLi/Naratiosateachofthesites. Forothercalculatedelemental ratios (i.e. Ca/Na, Na/Si, and Li/Si), correlations with runoff were not modeled with any regression equations. 3.4 Results 3.4.1 Atmospheric Inputs If the dissolution of Cl-bearing evaporite minerals was a major contributor to the riverine solute budget for any of the sampled catchments, the application of the measured Cl − concentrations to correct for atmospheric contributions would be compromised. The overall low concentrations of Cl- (2-15 μM, 1-12 μM, 6- 31 μM, and 6-60 μM for Way, SP, MLC and CICRA respectively; Tables A1-4) measured in the river samples and their similarity to the rainwater Cl − concen- trations determined in this study (0.4-11.5 μM for samples with Na/Cl − ratios near the seawater ratio) suggests negligible contribution of Cl − from evaporite dissolution, especially considering that the rainwater samples are dominantly from 40 high elevations (> 450 m) and Amazonian rainwater Cl − concentrations are known to increase with increasing proximity to the Atlantic ocean (Stallard & Edmond, 1981). Additionally, the relationship between element/Cl − and Na/Cl − ratios can be used to diagnose potential evaporite inputs. Marine evaporite deposits should have Na/Cl − ratios near 1, and thus would plot near the precipitation end-member (Na/Cl − = 0.86) and influence the calculated element/Cl- ratio based on the lin- ear regression model. Based on the data from Stallard (1974) and Gaillardet et al. (1997), the Ca/Cl − and Mg/Cl − ratios of plausible evaporite end-members for the Amazon system can be estimated to range from 0.06-0.26 and 0.008-0.03 respectively, for Na/Cl − ratios of 0.86-1. Because these ratios are offset from the calculated best-fit line from the data, and no data appear to trend towards these values, significant evaporite contribution can be discounted and Cl − concentration can be reliably used to correct for atmospheric inputs (Figure. 3.2). However, if Cl − bearing evaporites do contribute significantly to the dissolved budget of Cl − in any of the studied catchments, the correction procedure would have the dual effect of removing both atmospheric and evaporite contributions to the dissolved load. The calculated element/Cl − ratios for atmospheric inputs of K, Ca, Mg, and SO 4 2− as well as the assumed Na/Cl − ratio are all within the range of the ratios measured in precipitation samples collected as part of this study as well as those available from the literature (Stallard & Edmond 1981; Moquet et al. 2011; Fig- ure. 3.2). Between the different catchments, there is no significant variation in the composition of major element/Cl- ratios of atmospheric inputs. The average of the calculated K/Cl − (0.4±1), SO 4 2− /Cl − (1.5±5.5), Ca/Cl − (1.4±4.4), and Mg/Cl − (0.8±3) ratios are all higher than the seawater values but, because of the largeuncertaintiesassociatedwiththefittingprocedure,overlapwithinuncertainty (Figure. 3.2). Elevated ratios for these elements in precipitation may be expected if there is a small contribution from dissolution of dust particles or, in the case of K and SO 4 2− , biogenic aerosols. The calculated Sr/Cl − ratios are comparable to the ratio presented in Gaillardet et al. (1997), which was calculated using the data of Furch (1984). The calculated Li/Cl − ratios range from 0.0002-0.008 and are 1-2 orders of magnitude greater than the seawater Li/Cl − ratio (4.8×10 −5 ). The aver- age of the calculated Si/Cl − ratios of the atmospheric-input end member for all of the sites is 1.7± 5.8. This apparent enrichment of Si in rainfall may be related to 41 the dissolution of silicate mineral-rich dust (e.g., Reid et al. 2003) or the addition of solutes during the passage of rainfall through the forest canopy as throughfall (i.e. the addition of Si in preference to Cl − ; Tobón et al. 2004). Broadly, the frac- tional contribution of atmospheric deposition to the dissolved load at each of the study sites increases with decreasing mean catchment elevation due to an increase in Cl- concentrations. 3.4.2 Intra-site variation in elemental concentrations with runoff The observed concentration-runoff relationships show three different types of variation: 1) variation between different elements within the same catchment, 2) temporal variation of a single element within a single catchment, and 3) vari- ation between different catchments (Figure. 3.3-3.6, Table 3.2). The observed concentration-runoff relationships are first discussed on a catchment-by-catchment basis before comparing the results between catchments in the discussion section. In the Wayqecha catchment, all of the elements except for Cl − and K show systematic variation in concentration with runoff (i.e. adjusted R 2 > 0.3; Table 3.2) and the calculated values of the power law exponents are not significantly different between the atmospheric-input corrected and uncorrected samples (Table 3.2). Silicon has the lowest adjusted R 2 value (0.35; Figure. 3.3b; Table 3.2), but this is consistent with the fact that it has a near-zero power law exponent (-0.12± 0.03; Figure. 3.3b; Table 3.2) and is thus expected to vary little with runoff. Relative to Ca, Mg and Sr, higher adjusted R 2 values are observed for Na, Li, and SO 4 2− despite the fact that all of these elements have relatively similar power law exponents (-0.29 to -0.41; Figure. 3.3a,c-d; Table 3.2). Interestingly, the residualsofthepowerlawconcentration-runofffitsforCa,Mg, andSrarepositively correlated with the sample Sr/Na ratio while the residuals for the Na, Si, Li, and SO 4 2− concentration-runoff relationships are not. If the data are separated based on their Sr/Na ratio, then samples with low Sr/Na ratios (i.e. > 6.2 mmol/mol; dominant samples/fit) have lower power law exponents for Ca, Mg, and Sr relative to the total fit (Figure. 3.3d; Table 3.2). In the San Pedro catchment, all of the elements except for Cl − and K show systematic variation in concentration with runoff (i.e. adjusted R 2 > 0.3) and the 42 80 100 120 140 160 180 200 220 Na ( μM) 0 5 10 15 20 25 30 150 200 250 300 350 Runoff (mm/day) Si (μM) 1.5 2 2.5 3 3.5 Li (μM) 0.4 0.6 0.8 1 1.2 Sr (μM) Wet Wet-Dry Dry Dry-Wet Dominant Subordinate All data Way 3204 m - 50 km 2 - 25º 0 5 10 15 20 25 30 Runoff (mm/day) 0 5 10 15 20 25 30 Runoff (mm/day) 0 5 10 15 20 25 30 Runoff (mm/day) Figure 3.3: Wayqecha catchment (Way) concentration-runoff relation- ships. Concentration-runoff relationships for Na (panel A), Si (panel B), Li (panel C), and Sr (panel D) from the Way catchment. Samples are color coded with respect to their collection season. The best-fit trend line for the power law regres- sion using all of the data (i.e. the "total" fit) is shown as a dashed black line. Samples included in the dominant fit, which is presented as a solid black line, are presented as circles. Samples excluded from the dominant fit are presented as squares. calculated values of the power law exponents are not significantly different between the atmospheric-input corrected and uncorrected samples (Table 3.2). However, the power law fit to the total dataset systematically overestimates element con- centrations at the highest runoff rates (Figure 3.4a-d). To account for this, all the data at high temporal resolution during the wet season (see section 2.2) were fit with a separate power law relationship (i.e. subordinate fit, Figures 3.4a-d; Table 3.2). At low runoff (<12 mm/day; i.e. dominant fit), all elements other than Cl- and K have concentration-runoff relationships characterized by negative and near zero power law exponents (Figures 3.4a-d; Table 3.2). Between the different ele- ments, Si has the highest exponent value (-0.10± 0.05; Figure 3.4b). The elements SO 4 2− , Li, and Sr have similar power law exponent values near -0.3 while the ele- ments Na, Ca, and Mg have higher exponent values near -0.2 (Figures 3.4a,c-d). 43 At high runoff (>12 mm/day; i.e. subordinate fit), Si, Ca, Mg, Na, SO 4 2− and Sr all have power law exponent values near -1 whereas the exponent for Li is distinctly lower (-1.6± 0.5; Figures 3.4a-d; Table 3.2) and the exponent for K is distinctly higher (1.65± 0.5; Table 3.2). Additionally, the transition between the dominant and subordinate concentration-runoff relationships is marked by increases in the concentrations of SO 4 2− , K, and Li (Figure 3.4c). 0 5 10 15 20 50 100 150 Runoff (mm/day) Na ( μM) 150 200 250 Si (μM) 0.6 0.8 1 1.2 1.4 1.6 Li ( μM) 0.4 0.6 0.8 1 1.2 Sr (μM) SP - 2778 m - 161 km 2 - 27º 0 5 10 15 20 Runoff (mm/day) 0 5 10 15 20 Runoff (mm/day) 0 5 10 15 20 Runoff (mm/day) Wet Wet-Dry Dry Dry-Wet Dominant Subordinate All data Figure 3.4: San Pedro catchment (SP) concentration-runoff relationships. Concentration-runoff relationships for Na (panel A), Si (panel B), Li (panel C), and Sr (panel D) from the SP catchment. Samples are color coded with respect to their collection season. The best-fit trend line for the power law regression using all of the data (i.e. the "total" fit) is shown as a dashed black line. Samples included in the dominant fit, which is presented as a solid black line, are presented as circles. Samples included in the subordinate fit, which is presented as a solid grey line, are presented as squares. In the MLC catchment, all of the elements except for Cl − and K show sys- tematic variation in concentration with runoff (i.e. adjusted R 2 > 0.3; Table 3.2) and the calculated values of the power law exponents are not significantly differ- ent between the atmospheric-input corrected and uncorrected samples (Table 3.2). For Li, the total power law fit overestimates concentrations for samples collected during storms in the dry season (June-August) and dry-to-wet season transition 44 (October-November; Figure 3.5c). When investigated in more detail, these dry season and dry-to-wet season transition samples are also relatively depleted in Na, Ca, Mg, Sr, and Si, though to a lesser extent than observed for Li (Figure 3.5a-d). To account for this, separate power law fits were calculated for samples collected between December and May (the wet season and wet-to-dry season transition; i.e. dominant fit) and samples collected between June and November (the dry season and dry-to-wet season transition; i.e. subordinate fit; Figure 3.5). It is worth noting that 2 of the 19 samples collected between June and November have concentration-runoff relationships consistent with the wet season samples and were therefore not included in the subordinate regression. 0 5 10 15 20 50 100 150 Runoff (mm/day) Na ( μM) 100 150 200 250 Si (μM) 0.2 0.3 0.4 0.5 0.6 Li ( μM) 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Sr (μM) MLC 2012 m 6025 km 2 - 22º Wet Wet-Dry Dry Dry-Wet Dominant Subordinate All data 0 5 10 15 20 Runoff (mm/day) 0 5 10 15 20 Runoff (mm/day) 0 5 10 15 20 Runoff (mm/day) Figure 3.5: Manu Learning Center catchment (MLC) concentration- runoff relationships. Concentration-runoff relationships for Na (panel A), Si (panel B), Li (panel C), and Sr (panel D) from the MLC catchment. Samples are color coded with respect to their collection season. The best-fit trend line for the power law regression using all of the data (i.e. the "total" fit) is shown as a dashed black line. Samples included in the dominant fit, which is presented as a solid black line, are presented as circles. Samples included in the subordinate fit, which is presented as a solid grey line, are presented as squares. For the MLC subordinate regression, all elements except for Cl − and K have power law exponent values near -1 (Figures 3.5a-d; Table 3.2). For the MLC 45 dominant fit, the calculated power-law exponents are higher than during the dry season (Figures 3.5a-d; Table 3.2). For Si, the power law exponent is the highest (-0.25± 0.04) of all of the elements (Figure 3.5, Table 3.2). Relative to Si, the exponents are slightly lower for Na, Li, and Sr (-0.36± 0.08, -0.35± 0.07, and -0.46± 0.1 respectively; Figure 3.5a-d). The lowest exponent values are seen for Ca, Mg, and SO 4 2− (-0.46± 0.07, -0.47± 0.06, and -0.58± 0.09 respectively; Table 3.2). In the CICRA catchment, all of the elements except for Cl − show system- atic variation in concentration with runoff (i.e. adjusted R 2 > 0.3; Figure 3.6; Table 3.2). For Na, Mg, and Si, the power law exponents calculated from the atmospheric-input corrected data are significantly lower than the exponents cal- culated from the uncorrected data (Table 3.2). For all other elements, the cal- culated power law exponents are not significantly different between the corrected and uncorrected data (Table 3.2). The 3-hour resolution time-series sampling at CICRA reveals samples collected during the falling limb of wet season storms are relatively enriched in Ca, Sr, and Mg (Figure 3.6d). These samples showing ele- vated concentrations were excluded from the dominant regression. For both the dominant fit of the corrected and uncorrected data, K has the highest power law exponent (-0.45± 0.18 and -0.20± 0.08 respectively). Next to K, Si has the highest power law exponent for both the corrected and uncorrected data (-0.65 ± 0.21 and -0.4± 0.11 respectively). All of the other elements show power law exponents that are near -1 for the corrected data, with slightly higher values for the uncorrected data (Table 3.2). 3.4.3 Variations in elemental ratios with runoff Due to variations in the concentration-runoff relationships for different ele- ments within the same catchment, elemental ratios also vary with runoff. At the Wayqecha, San Pedro, and CICRA sites, Na/Si ratios show a broad decrease with increasing runoff (Figures 3.7c-3.10c). At the San Pedro site, the samples collected at the highest runoff rates show a slight increase in Na/Si, but the ratio remains significantly lower relative to samples collected at the lowest runoff rates (Figure 3.8c). At the MLC site, the variation in Na/Si is complicated in that some samples collected at high runoff rates have similar Na/Si ratios to those collected at low 46 0 2 4 6 8 10 100 150 200 Runoff (mm/day) Na ( μM) 0 2 4 6 8 10 150 200 250 300 Runoff (mm/day) Si ( μM) 0 2 4 6 8 10 0.1 0.15 0.2 0.25 0.3 0.35 Runoff (mm/day) Li ( μM) 0 2 4 6 8 10 1 2 3 4 5 Runoff (mm/day) Sr ( μM) Wet Wet-Dry Dry Dry-Wet Dominant Subordinate All data CICRA 822 m - 27830 km 2 - 9º Figure 3.6: CICRA catchment concentration-runoff relationships. Concentration-runoffrelationshipsforNa(panelA),Si(panelB),Li(panelC),and Sr (panel D) from the CICRA catchment. Samples are color coded with respect to their collection season. The best-fit trend line for the power law regression using all of the data (i.e. the "total" fit) is shown as a dashed black line. Samples included in the dominant fit, which is presented as a solid black line, are presented as circles. Samples excluded from the dominant fit are presented as squares. For the criteria used to select samples for the dominant fit, see the main text. runoff rates (Figure 3.5c). Similar to Na/Si ratios, Li/Si ratios also show a broad decrease with increasing runoff at all of the sites (Figures 3.7d-3.10d). At the San Pedro site, Li/Si ratios are distinct from other sites in that there are maximum values at both high and low runoff (Figure 3.8d). At the MLC catchment, Li/Si ratios are distinct in that there are minima at both high and low runoff (Figure 3.9d). In addition to Na/Si and Li/Si ratios, Sr/Na ratios also show systematic vari- ation with runoff in all of the catchments except Wayqecha (Figures 3.7b-3.10b). For the Wayqecha catchment, Sr/Na ratios of the low Sr/Na samples do not vary systematically with discharge as evidenced by slopes that overlap with zero within the 95% CI for both the atmospheric-input corrected (3.6×10 −6 ± 16×10 −6 ) and 47 0 5 10 15 20 25 30 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 Runoff (mm/day) Ca/Na 0 5 10 15 20 25 30 3 4 5 6 7 8 x 10 −3 Runoff (mm/day) Sr/Na 0 5 10 15 20 25 30 0.4 0.5 0.6 0.7 0.8 Runoff (mm/day) Na/Si 0 5 10 15 20 25 30 0.005 0.01 0.015 Runoff (mm/day) Li/Si Way - 3204 m - 50 km 2 - 25º Wet Wet-Dry Dry Dry-Wet Dominant Subordinate Figure 3.7: Wayqecha catchment (Way) ratio-runoff relationships. Ratio- runoff relationships for Ca/Na (panel A), Sr/Na (panel B), Na/Si (panel C), and Li/Si (panel D) from the Way catchment. Samples are color coded with respect to their collection season. In panel B, the best-fit trend line for the linear regression using all of the data (i.e. the "total" fit) is shown as a solid black line with the black dashed lines showing the 95% confidence interval. Samples included in the dominant fit are presented as circles. Samples excluded from the dominant fit are presented as squares. For the criteria used to select samples for the dominant fit, see the main text. uncorrected(4.2×10 −6 ±15×10 −6 )elementaldata. IncludingallofWayqechasam- ples produces a slight positive correlation between runoff and Sr/Na (i.e. slopes of 38×10 −6 ± 26×10 −6 and 36×10 −6 ± 25×10 −6 for the corrected and uncorrected data respectively; Figure 3.7b). For the San Pedro, MLC, and CICRA catchments, Sr/Na ratios show a slight negative correlation with runoff and have slopes that are significantly different from zero for both the atmospheric-input corrected and uncorrected elemental data (Figure 3.8b-3.10b). During the falling limb of wet- season storms in the CICRA catchment, Sr/Na ratios are slightly elevated relative to the rising limb for both the atmospheric-input corrected and uncorrected ele- mental data (Figure 3.10b). These falling limb samples also show elevated Ca/Na ratios relative to all other samples collected at CICRA (Figure 3.10a). At the other sites, Ca/NaratiosarepositivelycorrelatedwithSr/Naratios(Figures3.7a-3.10a). 48 0 5 10 15 20 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 Runoff (mm/day) Ca/Na 0 5 10 15 20 4 5 6 7 8 9 10 x 10 −3 Runoff (mm/day) Sr/Na 0 5 10 15 20 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 Runoff (mm/day) Na/Si 0 5 10 15 20 3 4 5 6 7 8 x 10 −3 Runoff (mm/day) Li/Si Wet Wet-Dry Dry Dry-Wet Dominant Subordinate SP 2778 m - 161 km 2 - 27º Figure 3.8: San Pedro catchment (SP) ratio-runoff relationships. Ratio- runoff relationships for Ca/Na (panel A), Sr/Na (panel B), Na/Si (panel C), and Li/Si (panel D) from the SP catchment. Samples are color coded with respect to their collection season. In panel B, the best-fit trend line for the linear regression using all of the data (i.e. the "total" fit) is shown as a solid black line with the black dashed lines showing the 95% confidence interval. Samples included in the dominant fit are presented as circles. Samples excluded from the dominant fit are presented as squares. For the criteria used to select samples for the dominant fit, see the main text. 3.5 Discussion The variation in solute chemistry with runoff presented in this study is first used to identify specific processes that give rise to the variability seen between ele- ments in each catchment. This discussion includes sections 3.5.1 and 3.5.2, where elemental ratio-runoff relationships are used to examine how changes in the con- tribution of solutes sourced from different lithologies and the precipitation of sec- ondary mineral phases affect concentration-runoff relationships respectively. Pos- sible mechanisms for the temporal variability in concentration runoff relationships observed at each of the sites (e.g., during storm events) are discussed in section 3.5.3. After the details of concentration-runoff and ratio-runoff relationships are discussed for each catchment, the concentration-runoff relationships for selected 49 0 5 10 15 20 0.8 1 1.2 1.4 1.6 1.8 Runoff (mm/day) Ca/Na 0 5 10 15 20 2 3 4 5 6 7 8 x 10 −3 Runoff (mm/day) Sr/Na 0 5 10 15 20 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 Runoff (mm/day) Na/Si 0 5 10 15 20 1 1.5 2 2.5 x 10 −3 Runoff (mm/day) Li/Si Wet Wet-Dry Dry Dry-Wet Dominant Subordinate MLC 2012 m - 6025 km 2 - 22º Figure 3.9: Manu Learning Center catchment (MLC) ratio-runoff rela- tionships. Ratio-runoff relationships for Ca/Na (panel A), Sr/Na (panel B), Na/Si (panel C), and Li/Si (panel D) from the MLC catchment. Samples are color coded with respect to their collection season. In panel B, the best-fit trend line for the linear regression using all of the data (i.e. the "total" fit) is shown as a solid black line with the black dashed lines showing the 95% confidence interval. Samples included in the dominant fit are presented as circles. Samples excluded from the dominant fit are presented as squares. For the criteria used to select samples for the dominant fit, see the main text. elements are compared between catchments in order to identify how coupled varia- tion in geomorphic and hydrologic processes leads to differences in solute chemistry and chemical weathering (section 3.5.4). In section 3.5.5, the implications of co- variation in the geomorphic and hydrologic controls on chemical weathering for the larger Andes/Amazon system are discussed. 3.5.1 Lithologiccontrolsonconcentration-runoffandratio- runoff relationships TheobservedvariationinCa/NaandSr/NaratioswithrunoffattheSanPedro, MLC, and CICRA sites (Figures 3.8-3.10) suggests that the relative contribution 50 0 2 4 6 8 10 1.5 2 2.5 3 3.5 4 4.5 Runoff (mm/day) Ca/Na 0 2 4 6 8 10 0 0.01 0.02 0.03 0.04 Runoff (mm/day) Sr/Na 0 2 4 6 8 10 0.4 0.5 0.6 0.7 0.8 0.9 Runoff (mm/day) Na/Si 0 2 4 6 8 10 0.6 0.8 1 1.2 1.4 1.6 x 10 −3 Runoff (mm/day) Li/Si Wet Wet-Dry Dry Dry-Wet Dominant Subordinate CICRA 822 m - 27830 km 2 - 9º Figure3.10: CICRAcatchmentratio-runoffrelationships. Ratio-runoffrela- tionships for Ca/Na (panel A), Sr/Na (panel B), Na/Si (panel C), and Li/Si (panel D) from the CICRA catchment. Samples are color coded with respect to their col- lection season. In panel B, the best-fit trend line for the linear regression using all of the data (i.e. the "total" fit) is shown as a solid black line with the black dashed lines showing the 95% confidence interval. Samples included in the dominant fit are presented as circles. Samples excluded from the dominant fit are presented as squares. For the criteria used to select samples for the dominant fit, see the main text. of silicate mineral dissolution to the dissolved load changes with runoff (Gaillardet et al., 1999; Tipper et al., 2006; Calmels et al., 2011). Typically, silicate rocks are characterized by lower Ca/Na and Sr/Na ratios relative to carbonate rocks (Gaillardet et al., 1999). Thus, the slope of the Ca/Na and Sr/Na ratio-runoff relationships at San Pedro, MLC, and CICRA imply that the relative proportion of solutes derived from silicate mineral dissolution increases with increasing runoff (Figures 3.8-3.10). Potentially, this behavior could result from the carbonate reac- tion front being deeper than the silicate reaction front, as observed in some settings (e.g., Brantley et al. 2013). As the water table rises above the carbonate reaction front at higher flow rates, the proportion of solutes derived from silicate weather- ing would increase (cf. Godsey et al. 2009). Regardless of the exact mechanism, 51 changes in the relative contributions of carbonate versus silicate weathering with runoff complicates the interpretation of the concentration-runoff relationships of Ca, Mg, and Sr because these elements are sourced from both silicate and car- bonate minerals (Gaillardet et al., 1999). As a result, the rest of the discussion will focus on Na, Li, and Si because these elements are sourced only from silicate mineral dissolution in the studied catchments. 3.5.2 The effects of secondary mineral formation on concentration-runoff and ratio-runoff relationships The power law exponent describing the relationship between Si concentrations and discharge is consistently higher than Na and Li at all sites despite the fact that all three elements have the same lithologic source (Figures 3.13-3.14; Table 3.2). TheconcentrationsofSialsoremainrelativelyconstantbetweenthesampling sites. Together, these observations suggest that Si concentrations are maintained by equilibrium with respect to a secondary silicate mineral, which is consistent with other hydrochemical studies (Clow & Mast, 2010; Maher, 2011). Indepen- dent evidence for near-equilibrium conditions comes from the observed variation in element/Si ratios with runoff (Figures 3.7c,d-3.10c,d). During silicate min- eral dissolution, both cations (e.g., Na and Li) as well as Si are released into the dissolved phase. As dissolution progresses, saturation with respect to secondary silicates can buffer the concentration of dissolved Si while the concentrations of dissolved cations that are not readily partitioned into secondary silicates (e.g. Na) continue to increase. The removal of Si but not cations should lead to an increase in the element/Si ratio (e.g., Georg et al. 2007). Within this simple framework, it is possible to interpret intra-catchment variations in Na/Si and Li/Si with runoff as partly reflecting the balance between primary silicate mineral dissolution and secondary mineral precipitation, because Na, Li, and Si are all sourced dominantly from silicate minerals (note that interpreting ratios with Ca, Mg, and Sr is more complicated because of carbonate sources; see section 3.5.1). The general observation that Na/Si and Li/Si ratios decrease with increasing runoff is consistent with the river composition at low runoff being dominated by inputsfromfluidswithlongtransittimes(i.e. groundwaters)thathaveexperienced a greater extent of Si removal as a result of secondary mineral formation (Figures 52 3.7c,d-3.10c,d). As runoff increases, fluids with shorter transit times that have not experienced the same extent of secondary mineral formation become more dominant and act to decrease the riverine Na/Si and Li/Si ratios (Figures 3.7c,d- 3.10c,d). 10 -3 10 -2 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Li/Na Na/Si 0.8 0.7 0.6 0.5 0.5 1 1.5 2 1000*Li/Na Na/Si CICRA MLC SP WAY Increasing elevation Clay formation with Li uptake Clay formation without Li uptake r 2 = 0.27 Figure 3.11: Geochemical behavior of Li. A comparison between Li/Na and Na/Si ratios between the sampling sites. (A) Log-Log comparison of all sites. (B) A focus on samples from CICRA plotted on a linear scale. A linear regression, 95% confidence interval, and R 2 value are shown for the relationship between Li/Na and Na/Si at CICRA Relative to Na, Li is more readily incorporated into secondary silicate phases (e.g., Huh et al. 2004b). As a result, comparisons of the variation in Na, Li, and Si concentrations in each catchment can yield additional insight into secondary mineral formation processes and also inform the use of Li as a paleo-weathering proxy (e.g. Misra & Froelich 2012). If Li behaves as conservatively as Na (i.e. if neither element is incorporated significantly into secondary silicates), then there should be no relationship between Na/Si and Li/Na ratios in a given catchment. If Li is significantly incorporated into secondary silicates relative to Na, then there should be a negative correlation between Na/Si and Li/Na ratios with a slope that reflects the affinity of Li for the secondary phase relative to Na. In the Andean catchments (SP and Way), Li behaves as conservatively as Na (Way; Figure 3.11) 53 or is only slightly incorporated into secondary silicates (SP; Figure 3.11). At the Andean outlet site (MLC), Li is slightly incorporated into secondary silicates (Figure 3.11). In the foreland floodplain (CICRA), Li is strongly incorporated into secondary phases relative to Na (Figure 3.11). Potentially, this behavior is related to a change in the composition of the dominant secondary silicate phase between the Andes and foreland-floodplain as suggested by downstream changes in the clay mineralogy of suspended sediments (Guyot et al., 2007). The changing behavior of Li relative to Na observed in the Madre de Dios system (this study) is consistent with a recent study of the adjacent Beni river, which shows Li incorporation into secondary minerals across the mountain-floodplain transition based on Li isotope evidence (Dellinger et al., 2014). 3.5.3 The effects of changes in fluid flow paths on concentration-runoff and ratio-runoff relationships While the general trends in Na/Si and Li/Si ratios with runoff are consistent with the conceptual framework presented in section 3.5.2 (i.e. a decrease in Na/Si and Li/Si with increasing runoff resulting from a greater extent of Si removal at low runoff / long fluid transit times), some notable deviations occur (e.g., Figures 3.7c,d-3.10c,d). In the San Pedro catchment, elevated Na/Si and Li/Si ratios are seen at both high and low runoff (Figures 3.8c & 3.8d). It is unlikely that this represents an additional input from a concentrated groundwater source at high runoff because the concentration-runoff relationships for most elements during this period suggest dilution (i.e. power law slopes of the subordinate fits near -1; Figure 3.4, Table 3.2). Dilution at high runoff may be caused by a shift towards fast, near surface flow paths with reduced water-rock interaction time, as suggested by Stallard & Murphy (2014) to explain the dilution in dissolved organic carbon at high runoff in streams in Puerto Rico, although they did not see the same dilution of other elements (e.g., Na and Si), which I see in the SP catchment. Nonetheless, a switch to fast, near-surface flow paths at SP is consistent with the observation from this site that the fraction of particulate organic carbon derived from recently living biomass (POC non−fossil ) also increases at high runoff rates, because the erosion of surface soils by fast, near surface flow would be expected to mobilize POC non−fossil (Clark et al., 2013, 2014). 54 The observation from the SP catchment that K concentrations increase with increasing runoff at runoff rates > 12 mm day −1 rates despite the overall trend of dilution for other solutes is also consistent with a shift to fast, near-surface flow paths . Potassium can be enriched in surface soil layers because it is present as an exchangeable cation and/or in biomass at relatively high concentrations (Elsenbeer et al., 1995; Boy et al., 2008). At high runoff, Li concentrations are also elevated (Figure3.4c), butapotentialsourceoftheLienrichmentishardtoidentifybecause Li is not taken up appreciably by plants (Lemarchand et al., 2010). Li could be derived from clay mineral dissolution and though this should also release Si, plant uptake may remove Si from soil solutions (Derry et al., 2005). It is also possible that Li is enriched in the surface-soil exchange pool and/or preferentially leached during mineral dissolution. Involvement of the surface-soil exchange pool is consistent with the rapid change in Li concentrations (i.e. exchange reactions occur rapidly; Clow & Mast 2010; Kim et al. 2014) and the fact that elevated Li concentrations are not sustained. Reactive transport through a uniform mineral substrate would not generate the complex behavior in concentration-runoff and ratio-runoff relationships seen at SP. Instead, it is likely that the dominant fluid flow paths at SP vary in response to runoff, as outlined schematically in Figure 3.12. Because this complex behav- ior occurs at high runoff rates, it has an important impact on the total solute fluxes. Compared to the other sites, San Pedro has the highest annual precipita- tion (Lambs et al., 2012; Clark et al., 2014), which may be a factor contributing to its unique behavior at high runoff. Similar to the SP catchment, the observed variation in Li/Si with runoff at MLC cannot be described by a single trend, as minima are present at both high and low runoff (Figure 3.9). At low runoff, the decrease in in the Li/Si ratio is associated with the dilution of Li during dry season storms (Figures. 3.5c and 3.9d). Dilution of Li would be expected if Li-bearing primary phases were depleted in the sediments accessed by dry season storm flow paths. This is consistent with the overall trend of decreasing Li concentrations with elevation, which suggests the preferential loss of Li-bearing phases. At high runoff, lower Li/Si ratios are consistent with shorter fluid residence times favoring less sequestration of Si in secondary phases (see section 3.5.2). 55 During the falling limb of wet season storms, samples from the CICRA catch- ment show elevated concentrations for some solutes (Figure 3.6) as well as elevated Ca/Na ratios (Figure 3.10a). This behavior could stem from either temporal vari- ability in the relative contributions of different tributaries or a catchment-scale change in the dominant fluid flow paths. While these possibilities cannot be dis- tinguishedwiththeavailabledata, thecoupledincreaseintheCa/Naratiosuggests increased inputs from the dissolution of carbonate minerals (Gaillardet et al. 1999, Figure 3.10a). Runoff Concentrations of bedrock-derived solutes Concentrations of biosphere-derived solutes and [POC] non-fossil High Runoff Near-surface flow paths dominant -leaching of biosphere-derived solutes (e.g. K) -enhanced erosion of POC non-fossil -dilution of bedrock-derived solute concentrations Low Runoff Deep flow paths dominant -”Chemostatic” behavior for bedrock-derived solutes -[Si] buffered by secondary mineral precipitation Andean Catchment Hysteresis Concentration Figure 3.12: Schematic of the hypothesized effects of flow path variability on Andean concentration-runoff relationships. At high runoff rates (i.e. subordinate samples/fit), samples collected at the San Pedro gauging station in the Andes reveal coupled changes in the concentration runoff relationships for multiple elements consistent with a switch to rapid, near-surface flow paths (see section 3.5.3). 56 At the Wayqecha site, the river chemistry shows a bimodal behavior with sam- ples showing two distinct modes of solute concentrations and Sr/Na ratios (Figures 3.3,3.7b). However, unlike the other sites, there appears to be no clear temporal or hydrologic control on the shift between the two distinct solute chemistries. The Wayqecha sampling locality is immediately downstream from a confluence of two streams. It is possible that the variation in the solute chemistry reflects changes in the mixing proportions of the two streams, because the west tributary typically shows higher Sr/Na ratios and solute concentrations. For example, the mixing proportions of the two streams may vary in response to changes in the spatial distribution of precipitation within the catchment, leading to variable trends in Sr/Na. 3.5.4 Geomorphic control on concentration-runoff rela- tionships in the Andes Amazon system The power law exponent that relates elemental concentrations and runoff in equation 3.1 for both the total and dominant fits decreases with decreasing mean catchment elevation and mean catchment slope angle (Figures 3.13-3.14). This behavior is true for all elements with concentrations that are highly correlated with runoff (i.e. excluding Cl- and K for most catchments) as well as for both the atmospheric-input corrected and uncorrected data (Figures 3.13-3.14). Focusing on Na and Si, because they are major elements sourced only from silicate mineral dissolution (see section 3.5.1), a linear function appears to describe the relationship between the power law exponents and mean catchment slope angle (Figure 3.14). For both the total and dominant fits, the exponents for Na and Si calculated for the Andean sites (Way and SP) do not overlap with the foreland floodplain site (CICRA) within their respective 95% confidence intervals (Figure 3.14, Table 3.2). Thus, although the relationships between the power law exponents and mean catchment slope angle are shallower for the total fits relative to the dominant fits (Figure 3.14), the Andean and foreland floodplain sites still show significant differences. The slope of this relationship is also shallower for the uncorrected data relative to the atmospheric-input corrected data, but this is largely due to the fact that the correction is significant for only the foreland floodplain site (Figure 3.14). 57 10 15 20 25 Power Law Exponent SO 4 2- Li Si Sr Na Ca Mg 10 15 20 25 −1.4 −1.2 −1.0 −0.8 −0.6 −0.4 −0.2 0 Power Law Exponent −1.4 −1.2 −1.0 −0.8 −0.6 −0.4 −0.2 0 Mean slope (degrees) Mean slope (degrees) Total concentrations Atmospheric input corrected concentrations Figure 3.13: Correlation between the calculated power law exponents and mean catchment slope angle for all solutes. Linear regressions of the power law exponents for the dominant fits of all solutes (except Cl- and K) against mean catchment slope angles. The observed correlation between the power law exponents and mean catch- ment slope angle, independent of how the data are corrected, implies that the hydrologic control on chemical weathering fluxes (i.e. the concentration-runoff relationship) is related to the geomorphic regime as reflected in the measurement of mean catchment slope angle. The observed range of the power-law exponents for Na and Si in the Andes-Amazon system (Figures 3.13-3.14, Table 3.2) is com- parable to the range seen in compilations of data from U.S. streams (Godsey et al., 2009) and global rivers (Moon et al., 2014). The surface areas of the catchments included in this study are generally greater than those in the U.S. stream compi- lation (Godsey et al., 2009) and smaller than those in the global river compilation (Moon et al., 2014). 58 5 10 15 20 25 30 −1 −0.8 −0.6 −0.4 −0.2 0 Mean Catchment Slope Angle (degrees) Power Law Exponent 5 10 15 20 25 30 −1 −0.8 −0.6 −0.4 −0.2 0 Mean Catchment Slope Angle (degrees) Power Law Exponent Si (primary fit) Si (total fit) Na (primary fit) Na (total fit) “Chemostatlc” b = 0 Dilution b = -1 A. Total concentrations B. Atmospheric input corrected concentrations Figure 3.14: Correlation between the calculated power law exponents and mean catchment slope angle for Na and Si. Calculated power law exponents for Na (magenta circles) and Si (black diamonds) plotted against mean catchment slope angle. Exponents calculated for the total fit are presented as open symbols. Exponents calculated for the dominant fit are presented as solid symbols. Also shown are linear regressions of the power law exponents for the total (dashed lines) and dominant fits (solid lines) against mean catchment slope angle. The results of this study show that the relationship between solute concentra- tions and runoff varies markedly between the Andean and foreland floodplain sites (Figures 3.13-3.14). Previous research in catchment hydrology and geochemistry has suggested multiple mechanisms that could produce concentration-runoff rela- tionships (e.g., Johnson et al. 1969; Christophersen et al. 1990; Godsey et al. 2009; Maher 2011; Maher & Chamberlain 2014). Of these models, the solute production model (Maher, 2011; Maher & Chamberlain, 2014) and the porosity-permeability- aperture model (PPA model; Godsey et al. 2009) produce functional forms that are consistent with field data and include internally consistent parameter values (God- sey et al., 2009; Maher, 2011; Maher & Chamberlain, 2014). While the available formulations of these models make many assumptions that may be inconsistent with natural systems, the underlying processes that they describe are likely to be important. As a result, qualitative comparisons between model predictions and field data can be useful. The relevant details of the solute production model and PPA model are included below and are then followed by a discussion of how key 59 model parameters may vary between the Andes-Amazon sites and give rise to the observed change in the power law exponent with the geomorphic regime. In the solute production model, the reaction kinetics and the abundance of primary reactive minerals set the timescale required for solute concentrations to approach the thermodynamic equilibrium concentration (i.e. the solute concentra- tion where either the primary mineral is in equilibrium with the fluid or there is a balance between the rates of primary mineral dissolution and secondary mineral precipitation; Maher 2011; Maher & Chamberlain 2014). If fluid transit times are long relative to the timescale required to approach equilibrium, the catchment will display chemostatic behavior (Maher, 2011; Maher & Chamberlain, 2014). If fluid transit times are short relative to the timescale required to approach equilibrium, then the catchment will show dilution. High rates of erosion can increase the abun- dance of primary reactive minerals by shortening mineral residence times (Gabet & Mudd, 2009; Ferrier & Kirchner, 2008; Hilley et al., 2010; West, 2012; Li et al., 2014; Maher & Chamberlain, 2014). While this means that fluids will quickly approach equilibrium in rapidly eroding environments, high solute fluxes will only be observed if fluid transit times are sufficiently long (Maher, 2011; Maher & Chamberlain, 2014). By creating relief, rapid rates of erosion may also contribute to increasing flow path lengths and fluid transit times (e.g., McGuire et al. 2005), though the general relevance of this effect is not known. If the porosity, permeability, and average pore aperture of the critical zone decays exponentially with depth, then the mineral surface area in contact with fluids will increase as the water table rises. Depending upon the length scale over which porosity, permeability, and average pore aperture decay, the increase in the available surface area with runoff will variably offset the increased water flux leading to a power-law scaling between concentration and runoff (Godsey et al., 2009). If the porosity, permeability, and pore aperture all decay over relatively shorter length scales (i.e. a heterogeneous critical zone), then the system will be more "chemostatic." If the porosity, permeability, and pore aperture all decay over relatively longer length scales (i.e. a more homogeneous critical zone), then the system will dilute to a greater extent with increasing runoff. However, as noted by Godsey et al. (2009), the depth-dependent variation in primary mineral abundances must also factor in as it also controls the mineral surface area available for reaction. 60 Fluid transit times as a possible control on concentration-runoff rela- tionships The systematic variation in the calculated power law exponent with mean catchment slope angle (Figures 3.13-3.14) suggests that an underlying control on the concentration-runoff relationship of each catchment is somehow related to topography and/or erosional regime. As is often the case in natural gradients, multiple environmental factors co-vary with mean catchment slope angle in the study area. One potential driver might be differences in fluid transit times (i.e. the time it takes rainfall to transit through the catchment and enter the stream channel) between the sites. Changes in the fluid transit time could modulate the concentration-runoff relationship by influencing the length of time available for solute concentrations to approach the thermodynamic equilibrium concentration (Maher, 2010, 2011; Maher & Chamberlain, 2014). At present, there are few direct constraints on fluid transit times in any of the studied catchments, but there is some relevant evidence from catchment water bal- ances and temporal variations in the water isotopic composition of rainwater and river flow. Analysis of the monthly water budgets of the Way and SP catchments suggests that precipitation delivered during the wet season (December-March) is transiently stored within the catchment and discharged during the dry season (May-September; Clark et al. 2014). While the isotopic composition of precipita- tion shows a large seasonal cycle (i.e. amplitude of approximately 70h in δD), the isotopic composition of the river remains relatively constant (i.e. amplitude of approximately 10h in δD; Clark et al. 2014). There exist many complex models for quantitatively interpreting this isotopic dampening in terms of mean fluid transit times (e.g., McGuire & McDonnell 2006), all indicate that the extent of dampening is inversely proportional to the relative contribution of water with short fluid transit times. Thus, the seasonal water budget and the large extent of isotopic dampening in the Way and SP catchments imply relatively long fluid transit times. Seasonal measurements of tributaries across the elevational gradient show that at lower elevations, the stream waters are more sensitive to seasonal changes in the isotopic composition of precipitation (i.e. less isotopic dampening at lower eleva- tion; Ponton et al. 2014). This implies that fluids with shorter transit times make 61 a greater contribution to stream flow at lower elevations (i.e. CICRA catchment). Although there is not enough data to constrain the mean and distribution of fluid transit times robustly, especially at CICRA, the existing data are consistent with a increase in the proportion of fluids with short transit times with decreasing mean catchment slope angle. This difference in fluid transit times may partly explain the observed decrease in the power law exponent with decreasing mean catchment slope angles (Figures 3.13-3.14). For the same net reaction rates, a greater relative contribution of fluids with shorter transit times in the foreland floodplain would result in greater solute dilution with increasing runoff (i.e. lower power law expo- nent), because the time available for reaction would be shorter (Maher, 2010, 2011; Maher & Chamberlain, 2014). Mineral residence times as a possible control on power law exponents Changes in concentration-runoff relationships between the sites may also be related to changes in mineral residence times (i.e. the duration of time minerals spend weathering in the critical zone before being exported by erosion). Because the Andes Mountains are the dominant source of sediments in the Amazon system (Gibbs, 1967), there is a timescale associated with the transport of material to and through the foreland floodplain regions (Dosseto et al., 2006; Bouchez et al., 2012). At the river basin scale, this timescale can be considered the residence time of minerals within the critical zone. As material is transported, it undergoes reaction and becomes increasingly depleted in primary reactive minerals. If the timescale associated with sediment transfer from the Andes through the foreland floodplain is sufficiently long, then this would lead to a depletion of primary reactive minerals and increase the time required for fluids to reach thermodynamic equilibrium in the foreland floodplain (Maher & Chamberlain, 2014). All else being equal, a longer time-to-equilibrium should result in greater solute dilution with increasing runoff (Maher & Chamberlain, 2014). Because fluids with short transit times are thought to make a greater contribution to river flow in the foreland floodplain (see section 3.5.4), the effect of an increase in the time required to reach equilibrium on the concentration-runoff relationship is amplified leading to even greater solute dilution. 62 Across a similar elevation gradient in the adjacent Bolivian Andes, U-series analyses suggest an increase in sediment residences times with decreasing elevation (Dosseto et al., 2006). If residence times vary with elevation in a similar fashion in the Kosñipata-Madre de Dios system and lead to depletion in primary mineral concentrations, then this could be a plausible mechanism to cause the observed change in the concentration-runoff relationship with elevation. Fluid flow paths and weathering zone thickness as a possible controls concentration-runoff relationships The thickness of the weathering zone is thought to influence the concentration- runoff relationship by defining the length scale over which reactions occur and, for a given flow rate, the fluid transit time (Maher, 2011; Maher & Chamberlain, 2014). Longer weathering length scales, or "deeper" weathering zones, would lead to higher power law exponents and favor "chemostatic" behavior (Maher, 2011; Maher & Chamberlain, 2014). Some models (e.g., Gabet & Mudd 2009) assume that weathering reactions only occur in soils and regolith and that the thickness of this weathering zone is inversely related to erosion rates (Lebedeva et al., 2007). In the Andes-Amazon system, erosion rates are higher in the Andes Mountains (Gibbs, 1967; Meade, 1985) and, globally, erosion rates are correlated with mean catchment slope angle (Portenga & Bierman, 2011; Larsen et al., 2014). Together, these observations would suggest that the weathering zone would be deeper in the foreland floodplain catchment relative to the Andean catchments leading to higher power law exponent values (i.e. "chemostatic" behavior) in the foreland floodplain catchment, which is the opposite of what is observed (Figures 3.13-3.14). In contrast to an inverse relationship between weathering zone thickness and erosion rates, global datasets are best explained if the thickness of the weather- ing zone does not vary considerably with erosion (West, 2012). If the depth of the weathering zone is relatively constant, a significant proportion of the weath- ering zone at high erosion sites with thin soils is expected to be within fractured bedrock. Indeed, hydrologic and hydrochemical studies from other regions con- clude that the depth of the hydrologically and chemically active weathering zone in steep mountain catchments, often dominated by sedimentary or metamorphic lithologies, extends into fractured bedrock (Anderson et al., 1997; Tipper et al., 63 2006; Calmels et al., 2011; Andermann et al., 2012; Salve et al., 2012; Clark et al., 2014;Kimetal.,2014). IntheAndeansitesinthisstudy(WayandSP),theannual water budget and water isotope mixing calculations suggest that significant vol- umesofwatertransitthroughfracturedbedrockbeforeenteringthestreamchannel (Clark et al., 2014). Thus, it is possible that some of the observed changes in the concentration-runoff relationships are related to the fact that a greater proportion of the flow path length in the Andean catchments is contained within fractured bedrock, which has a higher concentration of weatherable minerals than soils and regolith. By increasing the fraction of weatherable minerals, these bedrock flow paths would be expected to approach thermodynamic equilibrium more rapidly leading to chemostatic behavior (Maher, 2011; Maher & Chamberlain, 2014). The role of weathering in fractured bedrock (potentially with abundant primary min- erals) may be inherently related to increased fluid transit times (because of longer length scales), but it is not straightforward to disentangle the effects of reactive mineral availability verses fluid transit times based on the time-series data from this study. Targeted work to better understand the dominant flow paths that con- tribute solutes to river systems in the Andes-Amazon and other systems would be useful future contributions to the study of chemical weathering. Critical zone structure as a possible control on power law exponents In the framework of the PPA model, the observed decrease in the power law exponent with mean catchment slope angle could result from changes in the struc- ture of the critical zone. In the foreland floodplain, thick deposits of Andean sediments (10-50 m; Rigsby et al. 2009) may give rise to a critical zone that has a relatively homogenous pore volume structure such that changes in the depth of the water table do not lead to substantial changes in reactive surface area. Coupled with primary mineral depletion as a result of long mineral residence times (Dos- seto et al., 2006), these characteristics of the critical zone in the foreland floodplain would give rise to system where concentrations dilute with increasing runoff. In the Andean catchments, bedrock fractures may provide most of the reactive sur- face area (Clark et al., 2014). As a result, it is difficult to predict how the reactive surface area changes with water table depth in the Andean catchments. 64 3.5.5 Implications for the role of floodplain weathering in the Amazon system The role of the foreland floodplain in the overall Amazonian weathering budget has important implications for understanding of how weathering processes operate at a global scale (Bouchez et al., 2012). Currently, models that couple the effects of mineral dissolution kinetics and erosion within a framework of geomorphic steady state (e.g., Gabet & Mudd 2009; Hilley et al. 2010; West 2012) consider the time that minerals spend within upland, eroding catchments, but not the time spent in transit through lowland floodplains. Similarly, models linking fluid chemistry and erosion rates (e.g., Li et al. 2014; Maher & Chamberlain 2014) consider an eroding columndevelopedfrombedrock. Arguably,thetransportanddepositionofreactive sediments in floodplains constitutes a distinct physical mechanism where reactive material is supplied to the top of the weathering column as opposed to being supplied from below as is commonly modeled, although the dynamics of storage and re-working of sediments within floodplains adds complexity (e.g., Bradley & Tucker 2013). If the weathering flux supplied to the ocean as a result of mountain uplift is dominantly the result of material supply to floodplains, then additional factors need to be considered to model these effects over geologic timescales. The observation that the power law exponent relating concentration and runoff decreases with decreasing mean catchment slope angle has important implica- tions for weathering processes in the foreland floodplain region (Figures 3.13-3.14). Solute concentrations in the foreland floodplain region show more dilution than in theAndes, meaningthatthecorrespondingweatheringfluxesarerelativelyinsensi- tive to changes in runoff over seasonal timescales (Figures 3.3-3.6; Table 3.1). This contrasts with the Andean catchments, which show a near 1:1 relationship between flux and runoff, implying a high sensitivity to changes in hydrologic conditions (Figures 3.3-3.6; 3.1). This is broadly consistent with the hypothesis that rapidly erodingenvironmentsaremoresensitivetochangesinrunoffrelativetomoreslowly erodingenvironments(e.g., West2012;Maher&Chamberlain2014). However, itis worth stressing that the results characterize the sensitivity of chemical weathering fluxes to seasonal changes in runoff. While seasonal concentration-runoff relation- ships can be stationary over inter-annual and decadal timescales (Godsey et al., 2009; Eiriksdottir et al., 2013) and have previously been interpreted to provide 65 insight into the climatic control on chemical weathering reactions (e.g., Tipper et al. 2006; Eiriksdottir et al. 2013), the dataset does not technically constrain the climatic sensitivity (sensu Stocker et al. 2014) of chemical weathering fluxes in the Andes-Amazon system. The general model of supply vs. kineticallylimited weathering regimes suggests that when weathering fluxes are limited by the supply of material by erosion, they are less sensitive to changes in temperature (West et al., 2005; West, 2012) and runoff (West et al., 2005; West, 2012; Maher & Chamberlain, 2014). While the results provide no constraints on the temperature or erosional dependence of foreland floodplain weathering fluxes, the lack of a significant runoff dependence is consistent with supply limitation in the foreland floodplain (West et al., 2005; West, 2012; Maher & Chamberlain, 2014). 3.6 Conclusions The results of this study show that the relationship between solute concen- trations and runoff varies systemically with mean catchment slope angle across a mountain-to-floodplain transition in the Peruvian Amazon. These observations provide new empirical data indicating an important link between physical erosion processes and hydrology in controlling catchment-scale weathering fluxes (Li et al., 2014; Maher & Chamberlain, 2014). Within each of the studied catchments, variability in the concentration-runoff and ratio-runoff relationships suggests that changesinlithology, theextentofsecondarymineralprecipitation, and, potentially, the dominant fluid-flow paths play an important second-order role in controlling solute chemistry. It is likely that combined variation in fluid transit times and mineral residence times gives rise to the variation in concentration-runoff and ratio-runoff relation- ships observed across the Andes to foreland floodplain gradient (Figure 3.15). It remains out of reach with the current data to quantitatively separate the relative roles of fluid transit times and mineral residence times, and to understand to what extent these are inherently linked within the Andes-Amazon system. However, the observations from this study emphasize that more detailed information about fluid and mineral residence times, and how they relate, may be the key to unlocking the 66 next door in understanding weathering systems. One implication of the observed variation in the link between runoff and weathering fluxes is that the Andes are more sensitive to changes in runoff than the foreland floodplain region (Figures 3.13, 3.14, 3.15). This is consistent with previous global models and suggests an important role for active erosion in global climate-weathering feedbacks, and in the global weathering thermostat (West, 2012; Maher & Chamberlain, 2014). 67 b exponent Total Concentrations SO 2− 4 ±(2σ) Li ±(2σ) Si ±(2σ) Sr ±(2σ) Wayqecha total -0.35 0.05 -0.41 0.05 -0.12 0.03 -0.29 0.06 Wayqecha dominant -0.38 0.04 -0.46 0.04 -0.13 0.03 -0.35 0.03 San Pedro total -0.22 0.04 -0.21 0.05 -0.14 0.03 -0.32 0.03 San Pedro dominant -0.27 0.08 -0.31 0.09 -0.10 0.05 -0.28 0.08 San Pedro subordinante -1.08 0.37 -1.60 0.50 -0.86 0.17 -1.09 0.32 MLC total -0.55 0.10 -0.30 0.13 -0.23 0.06 -0.44 0.12 MLC dominant -0.58 0.09 -0.36 0.08 -0.25 0.04 -0.47 0.10 MLC subordinante -1.59 0.53 -1.87 0.68 -0.75 0.31 -1.54 0.51 CICRA total -0.69 0.12 -0.56 0.14 -0.29 0.09 -0.90 0.23 CICRA dominant -0.86 0.19 -0.75 0.27 -0.40 0.12 -1.15 0.54 b exponent Total Concentrations Na ±(2σ) Ca ±(2σ) Mg ±(2σ) Wayqecha total -0.33 0.04 -0.30 0.05 -0.34 0.05 Wayqecha dominant -0.37 0.03 -0.34 0.04 -0.38 0.05 San Pedro total -0.23 0.02 -0.26 0.02 -0.24 0.02 San Pedro dominant -0.23 0.05 -0.20 0.04 -0.20 0.04 San Pedro subordinante -0.85 0.13 -0.98 0.17 -0.98 0.21 MLC total -0.31 0.09 -0.43 0.09 -0.43 0.09 MLC dominant -0.35 0.07 -0.46 0.07 -0.47 0.06 MLC subordinante -1.33 0.40 -1.37 0.40 -1.42 0.40 CICRA total -0.55 0.08 -0.43 0.13 -0.59 0.10 CICRA dominant -0.61 0.15 -0.64 0.13 -0.74 0.14 b exponent Corrected Concentrations SO 2− 4 ±(2σ) Li ±(2σ) Si ±(2σ) Sr ±(2σ) Wayqecha total -0.36 0.05 -0.42 0.05 -0.13 0.04 -0.30 0.06 Wayqecha dominant -0.39 0.04 -0.46 0.04 -0.14 0.03 -0.36 0.03 San Pedro total -0.24 0.04 -0.21 0.05 -0.15 0.03 -0.34 0.03 San Pedro dominant -0.30 0.09 -0.32 0.09 -0.12 0.06 -0.30 0.08 San Pedro subordinante -1.20 0.36 -1.65 0.52 -0.95 0.19 -1.23 0.36 MLC total -0.63 0.12 -0.30 0.13 -0.24 0.08 -0.47 0.13 MLC dominant -0.66 0.10 -0.36 0.08 -0.25 0.07 -0.49 0.11 MLC subordinante -1.01 1.95 -1.74 1.78 -1.06 1.05 -1.33 1.64 CICRA total -0.83 0.16 -0.72 0.19 -0.46 0.15 -1.07 0.29 CICRA dominant -1.07 0.30 -0.75 0.27 -0.65 0.21 -1.40 0.73 b exponent Corrected Concentrations Na ±(2σ) Ca ±(2σ) Mg ±(2σ) Wayqecha total -0.35 0.04 -0.31 0.05 -0.35 0.06 Wayqecha dominant -0.38 0.03 -0.36 0.04 -0.40 0.05 San Pedro total -0.25 0.02 -0.28 0.02 -0.26 0.03 San Pedro dominant -0.26 0.05 -0.22 0.04 -0.23 0.05 San Pedro subordinante -0.95 0.14 -1.10 0.16 -1.11 0.20 MLC total -0.34 0.09 -0.49 0.10 -0.48 0.10 MLC dominant -0.37 0.08 -0.51 0.08 -0.51 0.07 MLC subordinante -1.11 1.06 -1.23 1.25 -1.26 1.42 CICRA total -0.75 0.15 -0.50 0.16 -0.76 0.15 CICRA dominant -0.89 0.26 -0.77 0.17 -1.02 0.22 Table 3.2: Solute power-law exponents 68 -1.0 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 Flux Runo 1:1 (Solute ux = runo rate) Power Law exponent Fluid transit times Mineral residence times Short Long Short Long Foreland oodplain: far from chemostatic Long but variable mineral residence time -low erosion rate -thick soils depleted in reactive minerals Shorter uid transit times? -tributary δD responsive to changes in precipitation δD Andes Mountains: near “chemostatic” Short mineral residence time -weathering zone within fractured bedrock -high erosion rate -abundant reactive minerals Longer uid transit times -long owpaths through bedrock fractures -river and tributary δD not responsive to changes precipitation δD Runo Concentration -1.0 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 Power Law exponent Figure 3.15: Schematic of the coupled variation in fluid transit times and mineral residence times and their effects on concentration-runoff relationships across the mountain-to-floodplain transition. Together, the combined variation in fluid and mineral residence times determine the sensitivity of solute concentrations and fluxes to runoff. Contours of different power law expo- nents spanning from dilution (-1) to "chemostasis" (0) illustrate this variable sensi- tivity for arbitrary values of solute concentrations and runoff rates. For reference, the ranges in power law exponents measured for the Andean sites (blue) and the forelandfloodplainsite(green)areshownalongsidetheschematicofconcentration- runoff and flux-runoff relationships. 69 Chapter 4 Geomorphic influence on the acid budget of weathering in the Andes-Amazon system 4.1 Linking Statement This chapter reports the results of a field study aimed at quantifying the pro- portions of carbonic and sulfuric acids involved in chemical weathering reactions across a geomorphic gradient. As discussed in chapter 3, geomorphic gradients reflect changes in the characteristic timescales over which weathering reactions occur. While important to chemical weathering fluxes (Chapter 3 and Maher 2010, 2011; West 2012; Li et al. 2014; Maher & Chamberlain 2014), changes in fluid transit times and mineral residence times should also affect the extent to which trace reactive phases dominate chemical weathering budgets. While this hypothesis has been explored extensively for effect of trace calcite (e.g., Li et al. 2014), less research has been devoted to understanding the controls on trace sul- fide oxidation in natural systems despite its importance in biogeochemical cycles (Calmels et al., 2007) and the inception of chemical weathering (Brantley et al., 2013). This chapter addresses this knowledge gap by applying a suite of elemental and isotopic tracers to samples from the Peruvian Andes and Amazon. Of all of the chapters in this dissertation, chapter 4 involved the most collab- orative work. Samples were collected by myself, Dr. Kathryn Clark, Dr. Sarah Feakins, Dr. Camilo Ponton, Dr. Valier Galy, and Dr. Joshua West. Many of the water isotopic analyses were performed by Dr. Camilo Ponton in Alex Sessions’ lab at Caltech. All the Sr isotope analyses were performed by Dr. Julien Bouchez at IPGP. Dr. William Berelson allowed use of his Picarro for theδ 13 C measurements, which was operated by Nick Rollins. Dr. Jess Adkins allowed use of his laboratory 70 and Neptune MC-ICP-MS at Caltech for the dissolved sulfate-S isotope analyses, which were supervised by Dr. Guillaume Paris. 4.2 Introduction The interaction between water and rock at the Earth’s surface plays an impor- tantroleintheglobalbiogeochemicalcyclingofcarbon, sulfur, andoxygen(Walker et al., 1981; Berner et al., 1983; Berner, 1989). It has long been recognized that base-cation rich silicate minerals neutralize atmospherically-sourced carbonic acid, which drives the removal of carbon from the ocean-atmosphere system by provid- ing the cations and alkalinity necessary to precipitate and bury carbonate minerals (Walker et al., 1981; Berner et al., 1983). At the same time, minerals that are rich in reduced sulfur (e.g. pyrite; FeS 2 ) react with atmospheric oxygen (O 2 ) to pro- duce sulfuric acid (H 2 SO 4 ), which can dissolve carbonate minerals and liberate buried carbon into the ocean-atmosphere system (Calmels et al., 2007; Torres et al., 2014). The capacity of rocks to either release or remove carbon and oxygen from the ocean-atmosphere system is thus set by the relative contents of silicate, carbonate, and sulfide minerals, and their reactivity at Earth surface conditions (Brantley et al., 2013). While the supply of silicate, carbonate, and sulfide minerals is strongly depen- dentuponlithology, intrinsicvariabilityinthereactionratesofdifferentmineralsin surface environments can influence the actual net transfer of carbon during weath- ering. For example, sulfide minerals, which are often present as trace constituents of bedrock, can disproportionately affect weathering budgets due to their reactiv- ity (Hercod et al., 1998; Calmels et al., 2007). Similar effects are seen for trace carbonate phases (e.g., White et al. 1999). As a result, erosional and hydrologic processes, which together control the characteristic timescales over which weather- ing reactions occur (Maher & Chamberlain, 2014), should also influence the carbon budget of weathering reactions by modulating the extent to which highly reactive trace phases dominate the overall weathering budget. While the available data are consistent with this conceptual model (e.g., Calmels et al. 2007), few stud- ies have investigated the effects of changing acid sources on the carbon budget of weathering across environmental gradients. 71 In this study, I partition the dissolved load of rivers draining the Andes Moun- tains and Amazon foreland floodplain in Peru between the different mineral (i.e. carbonate and silicate minerals) and acid (i.e. carbonic and sulfuric acids) sources of solutes in order to determine the net transfer of carbon as a result of chemical weathering. The mixing model that is utilized incorporates measurements of major and trace element ratios, strontium isotope ratios, sulfur isotope ratios, and carbon isotope ratios measured on both solid and dissolved phase samples. I focus on the transition from the steep slopes and rapid erosion in the Andes to the depositional setting of the foreland floodplain of Peru because it is thought to reflect a gradient in the timescales that minerals undergo chemical weathering (Dosseto et al., 2006). Since the large rivers in the foreland floodplain region incorporate many distinct tributaries, I develop a tributary mixing model (c.f. Bickle et al. 2003) to quantify the contributions of all major tributaries to the mainstem under both wet and dry season conditions. 4.3 Methods 4.3.1 Study Site I focus on the Madre de Dios River, its major tributaries, and its headwaters in the Andean Kosñipata valley (Figure 3.1). The Madre de Dios river is a major tributary of the Madeira river, which is the second largest Amazonian river in terms of sulfate and bicarbonate fluxes (Gaillardet et al., 1997; Tardy et al., 2005). This renders quantitative knowledge of the carbon and sulfur sources in the head- water tributaries of the Madiera, such as the Madre de Dios River, important for understanding the total carbon and sulfur budgets of the Amazon system. Broadly, the rivers sampled for this study can be grouped based on the eleva- tion of their outlet and the percentage of the catchment’s contributing area located within the Andes Mountains (defined as elevations greater than 400 m.a.s.l follow- ing Moquet et al. 2011). Rivers with outlet elevations greater than 1300 m are termed high Andean rivers (Figure 3.1). Rivers with outlet elevations less than 1300 m and greater than 400 m are termed low Andean rivers (Figure 3.1). Rivers with outlet elevations less than 400 m are termed foreland-floodplain rivers (Figure 3.1). Foreland-floodplain rivers can either be pure foreland floodplain rivers (100% 72 contributing area < 400 m elevation) or mixed foreland-floodplain rivers (some contributing area > 400 m elevation). ThestudyofhighAndeanriversisfocusedaroundtheKosñipataRiversampled at the Wayqecha and San Pedro gauging stations (Figure 3.1). The catchment area of the Kosñipata River is predominately underlain by Paleozoic metasedimentary rocks with a minor proportion of felsic plutonic rocks. The study of low Andean rivers is focused around the Alto Madre de Dios river sampled at the Manu Learn- ing Center (MLC) gauging station (Figure 3.1). The catchment area of the Alto Madre de Dios river is predominately underlain by Paleozoic metasedimentary rocks, felsic plutonic rocks, and Paleozoic marine sedimentary rocks. The study of foreland floodplain rivers is focused around the Madre de Dios river sampled at the CICRA gauging station (Figure 3.1). In the foreland floodplain region, the catchment area of the Madre de Dios river is predominately underlain by sedi- ments shed from the Andes with additional contributions from Cretaceous marine sediments and Cenozoic continental deposits. In addition to year-long time series sampling of the mainstem rivers at each of the gauging stations, which is described in Clark et al. (2013), Clark et al. (2014), and Torres et al. (2015), tributaries within each of the catchments were sampled between 2010 and 2013 providing a wet and dry season pair for analysis (Ponton et al., 2014). Similarly, the mainstem of the Madre de Dios river downstream of the CICRA gauging station was sampled during both the dry and wet seasons in 2013 at Puerto Maldonado. Seasonal samples were also taken from the major tributaries that join the Madre de Dios river between the sampling localities at CICRA and Puerto Maldonado. The sampling locations and catchment characteristics for each of the localities are included in Table 3.1. Thisstudyutilizesawidevarietyofsampletypesthatwerecollectedatdifferent times and for different purposes (see below). As a result, not all of the geochemical analyses described below were preformed on every sample. As each analysis is introduced below, the number of samples analyzed is indicated in parenthesis. Discussion will focus on the core set of 37 dissolved samples that were analyzed for every dissolved phase geochemical parameter reported in this study. 73 4.3.2 Solid phase geochemical analyses Samples of river bank sediments and rocks were collected from across the entire study site in order to constrain the elemental and isotopic composition of differ- ent lithologic end-members. The details of each analytical method employed are included as supplementary material. Briefly, ground samples were used to measure the content of non-volatile elements using X-ray fluoresence spectroscopy (XRF; n = 25), the content and carbon isotopic composition of carbonate using infrared spectrometry on evolved CO 2 (IS; n = 16), and the sulfur isotopic composition of reduced sulfur compounds using an isotope ratio mass spectrometer (IRMS) after chromium extraction (n = 3; Gröger et al. 2009). Samples of un-ground riverbank sediments were also subjected to a sequential leachingprocedure toisolatethe chemical compositionofthe carbonatecomponent (Leleyter & Probst, 1999; Jacobson et al., 2003). The leachates were analyzed by microwave plasma atomic emission spectroscopy (MP-AES; n = 18). All of the analyzed leachates contain Al and Si, which suggests that both carbonate and silicate minerals are dissolved during the leaching procedure. To correct for this, a mixing model was developed that uses the measured Al/Ca ratios of the leachates and riverbed sediments to infer the Mg/Ca, Sr/Ca, and Na/Ca ratios of the carbonate component assuming that it contains no Al (see section D). 4.3.3 Dissolved phase geochemical analyses Water samples were collected using slightly different methods depending upon thesamplingyear. Regardlessofthesamplingyear,allsampleswerefilteredusinga 0.2μm filter within 24 hours of collection and split into multiple aliquots of which one was preserved by adding concentrated acid. The details of each analytical method employed are included as supplementary material. Briefly, the acidified samples were analyzed for cation and Si concentrations by MP-AES (n = 301), the sulfur isotopic composition of sulfate by multicollector inductively coupled plasma mass spectrometry (MC-ICP-MS) after chromatographic purification (n = 75; Paris et al. 2013), and the isotopic composition of Sr by MC-ICP-MS after chromatographic purification (n = 37). Un-acidified samples were analyzed for Cl and SO 4 concentrations by ion chromatography (n = 301), the concentration and isotopic composition of dissolved inorganic carbon (DIC) by IS on evolved CO 2 (n 74 = 81), and the isotopic composition of hydrogen and oxygen in water by IS (n = 24). For the DIC and water isotope measurements, only samples stored in glass exetainers without any head space were utilized. 4.3.4 Tributary mixing model During research cruises in March and August 2013, samples of the mainstem of the Alto Madre de Dios and Madre de Dios rivers were taken from both upstream and downstream of each major tributary confluence between the time series sam- pling stations at MLC and CICRA (Figure 4.1). At the same time as the mainstem was sampled, a sample of each tributary was also taken. This sampling approach allows for the relative contributions of the Alto Madre de Dios, Manu, Blanco, Chirbi, and Colorado rivers to the total water flux of the Madre de Dios river at CICRA to be estimated with a single conservative tracer (e.g., Cl − or δD). The details of this mixing calculation are described in Appendix E. By adding con- straints on the absolute water fluxes (i.e. discharge; m 3 /sec) of the Alto Madre de Dios river at MLC and the Madre de Dios river at CICRA at the time of sam- pling by using either measurements from an acoustic doppler current profiler or concentration-discharge relationships (Torres et al., 2015), the calculated relative contribution of each tributary to the total flux was used to estimate its discharge. During both sampling cruises, storm events near the gauging station at MLC occurred. For this reason, samples of the mainstem at MLC as well as downstream of the confluence between the mainstem and the Manu river, which is the first major tributary to join the mainstem past the MLC gauging station, were taken during both storm and background conditions. 4.4 Results 4.4.1 Solid phase analyses The riverbank sediments and rocks analyzed in this study are composed of a mixture of different silicate and carbonate minerals. In order to use the chemical composition of these solid phase samples to infer the lithologic sources of different solutes (see sections 4.5.1 & 4.5.2), it is necessary to determine the chemistry of 75 −72˚ −70˚ −12˚ 100 km Manu River Alto Madre de Dios at MLC Colorado River Madre de Dios River at CICRA Blanco River Chirbi River Figure 4.1: Tributary Map. Map showing the locations and catchment areas for the tributaries utilized in the mixing model. Together, the Alto Madre de Dios at MLC, Manu, Blanco, Chiribi, and Colorado River catchments account for 90% of the total catchment area of the Madre de Dios River at CICRA the carbonate and silicate components. Consequently, only the results of the solid phase analyses relevant to isolating the chemical composition of the carbonate and silicate phases are presented below. In the following sections, ratios with the subscript "sediment" refer to the XRF measurements of the bulk sediments. Ratios with the subscript "leach" refer to the analyses of leachates of the bulk sediments. Solid phase major and trace element ratios Bulk Na/Ca sediment , Mg/Ca sediment , and Sr/Ca sediment ratios range from 0.83- 36.71 mol/mol, 0.07-15.77 mol/mol, and 3.6-55.96 mmol/mol respectively (Figure 4.2). For all samples, the Na/Ca sediment and Sr/Ca sediment ratios are positively correlated. The relationship between Mg/Ca sediment ratios and both Na/Ca sediment and Sr/Ca sediment ratios is more complex and suggestive of two sample groups: a high Mg/Ca sediment group and a low Mg/Ca sediment group. 76 Measured carbonate contents range from 3.5 to 7080 μg carbonate carbon/g sediment. For some of these samples, the measured carbonate contents are suffi- cient to account for a significant proportion of the bulk Mg and Ca contents of the samples as measured by XRF. However, excluding samples where the molar car- bonate content is greater than 5% of the combined molar Ca and Mg content does not affect the ranges of measured Na/Ca sediment and Sr/Ca sediment ratios. Remov- ing the high carbonate samples slightly increases the lower bound of the range of Mg/Ca sediment ratios from 0.07 to 0.1 mol/mol. The raw riverbank leachates have Mg/Ca leach , Sr/Ca leach , Li/Ca leach , and Na/Ca leach ratios that range from 0.1-0.6 mol/mol, 1.8-14.5 mmol/mol, 1-42 mmol/mol, and 0.01-0.6 mol/mol respectively (Figure 4.2). After correcting for the effects of silicate mineral dissolution (see Appendix D), calculated Mg/Ca and Sr/Ca ratios of the carbonate component range from 0.01-0.3 mol/mol and 2-6 mmol/mol respectively (Figure D.1). For Na/Ca and Li/Ca ratios, the correc- tion procedure produced calculated ratios of the carbonate component that are not significantly different from zero. However, this is not unexpected since the uncertainty induced by the correction procedure is greater than typical carbonate Na/Ca and Li/Ca ratios (Gaillardet et al., 1997; Dellinger et al., 2015). Solid phase reduced inorganic sulfur isotope ratios The δ 34 S values of the three rock samples range from -5 to 0h. For these samples, there is no correlation between the pyrite content of the rock, which was measured as the mass of the extracted Ag 2 S, and the sulfur isotope ratio. Replicate sample extractions showed no variation above the analytical uncertainty of the measurement (± 0.2h). Solid phase carbonate-carbon isotope ratios For most samples, theδ 13 C values of PIC measured after sequential phosphoric acid extraction at 25 ◦ C and 60 ◦ C are not significantly different and range from -3 to -19h. For samples where there is a slight offset (e.g., R1940-190511), the δ 13 C value measured after extraction at 60 ◦ C is lower. Combining the results of the sequential extractions at 25 ◦ C and 60 ◦ C and comparing them to the measured 77 10 −2 10 −1 10 0 10 1 10 2 10 −2 10 −1 10 0 10 1 10 2 Na/Ca (mol/mol) Mg/Ca (mol/mol) 10 −1 10 0 10 1 10 2 10 −2 10 −1 10 0 10 1 10 2 Sr/Ca (mmol/mol) Mg/Ca (mol/mol) Acetic acid leachates Riverbank sediments (high carbonate) Riverbank sediments (low carbonate) Gaillardet et al. (1997) carbonate end-member Gaillardet et al. (1997) silicate end-member Gaillardet et al. (1997) carbonate end-member Gaillardet et al. (1997) silicate end-member A B Figure 4.2: Solid phase geochemistry. (A) Na/Ca and Mg/Ca ratios of the riverbank sediment samples (triangles) and carbonate leachates (squares). River- bank sediments with low carbonate contents are colored black. The measured values are compared to the end-members reported by Gaillardet et al. (1997). (B) Sr/Ca and Mg/Ca ratios of the riverbank sediment samples (triangles) and car- bonate leachates (squares). Riverbank sediments with low carbonate contents are colored black. The measured values are compared to the end-members carbonate contents reveals that samples with the highest carbonate contents have δ 13 C values near -5h. 4.4.2 Dissolved phase analyses For the time series samples taken from the mainstem rivers at Wayqecha, San Pedro, MLC and CICRA, variations in solute concentrations with discharge are discussed in Torres et al. (2015). Here, I focus on elemental and isotopic ratios for both the time series and tributary samples (see section 4.3.1) and consider their implications for acid sources and weathering reactions. In general, the elemental and isotopic ratios of the tributary samples encompass and are more variable than the elemental ratios of the mainstem river into which they drain. So, for simplicity, downstream changes in the composition of the mainstem rivers will be presented before the data on the tributaries. The data from each of the mainstem rivers is introduced in the downstream order of the river catchment outlets (from high elevation to low elevation). Note that the San Pedro river is a tributary of the Kosñipata River at San Pedro. 78 Dissolved phase major and trace element ratios The Kosñipata River at Wayqecha and San Pedro as well as the Alto Madre de Dios river at MLC have very similar ranges of Na/Ca ratios (0.49-0.76, 0.58-0.74, and 0.55-1.06 mol/mol respectively; Figure 4.3a). In contrast, the Madre de Dios river at CICRA has lower Na/Ca ratios relative to the Andean catchments (0.26- 0.52 mol/mol). An analogous trend is seen for Sr/Ca ratios with similar values between the Andean catchments (3.2-4.9, 3.2-5.6, 3.3-5.0 mmol/mol for the Kosñi- pata river at Wayqecha, the Kosñipata river at San Pedro, and the Alto Madre de Dios river at MLC respectively), but higher values for the foreland floodplain catchment (the Madre de Dios river at CICRA; 3.6-9.3 mmol/mol; Figure 4.3b). For Mg/Ca, the ranges decrease steadily from the highest elevation Andean catch- ment (i.e. the Kosñipata river at Wayqecha; 0.56-0.89 mol/mol) to the foreland floodplain catchment (i.e. the Madre de Dios river at CICRA; 0.22-0.32; Figure 4.3a). Both Li/Ca and SO 4 /Ca ratios are positively correlated with Mg/Ca ratios (Figure 4.3c-d). All of the tributaries of the Kosñipata River at Wayqecha and San Pedro have similar Sr/Ca ratios to the mainstem (Figure 4.3b). However, for Li/Ca, Mg/Ca, and Na/Ca, there are three distinct groups that differ in composition from the Kosñipata river. Group one shows higher Li/Ca, Mg/Ca, and Na/Ca ratios than the Kosñipata river (i.e. tributaries R2832 and R2300; Figure 4.3a-c). Group two shows lower Li/Ca and Mg/Ca ratios than the Kosñipata river (i.e. the San Pedro river; Figure 4.3a,c). Group three shows higher Mg/Ca but lower Li/Ca and Na/Ca ratios relative to the Kosñipata river (i.e. tributaries R2044 and R2050; Figure 4.3a-c). Tributaries of the Alto Madre de Dios river at MLC have a very similar elemen- tal ratio composition to the mainstem with the exception of the La Union river, which is more similar to the Kosñipata river. Most of the tributaries of the Madre de Dios river at CICRA have either similar or higher Na/Ca, Mg/Ca and Li/Ca ratios relative to the mainstem (Figure 4.3a-c). The only exception is the Manu river, which has the lowest Na/Ca, Mg/Ca, and Li/Ca ratios of all of the rivers studied (Figure 4.3a-c). 79 10 −1 10 0 10 −3 10 −2 Sr/Ca Na/Ca 10 −1 10 0 10 −1 10 0 Na/Ca Mg/Ca 10 −4 10 −3 10 −2 10 −1 10 −1 10 0 Li/Ca Mg/Ca 10 −4 10 −3 10 −2 10 −1 10 −2 10 −1 10 0 10 1 Li/Ca SO 4 /Ca Way SP MLC CICRA mainstem tributary A B C D Figure 4.3: Dissolved phase geochemistry. In all panels, Purple symbols refer to samples from the Kosñipata River at Wayqecha catchment, red symbols refer to samples from the Kosñipata River at San Pedro catchment, Blue symbols refer to samples from Alto Madre de Dios River at MLC catchment, green symbols refer to samples from Madre de Dios River at CICRA catchment. Large filled diamonds are samples from a mainstem river and small open diamonds are samples from a tributary. (A)Na/CaandMg/Caratiosofthedissolvedphasesamples. (B)Na/Ca and Sr/Ca ratios of the dissolved phase samples. (C) Li/Ca and Mg/Ca ratios of the dissolved phase samples. (D) Li/Ca and SO 4 /Ca ratios of the dissolved phase samples. Radiogenic strontium isotope ratios The individual analytical uncertainty on each of the dissolved 87 Sr/ 86 Sr ratio measurements is less then one permil. Nonetheless, given the large range of vari- ability between samples, the measured 87 Sr/ 86 Sr ratios will only be reported to three decimal places in the main text. Broadly, the 87 Sr/ 86 Sr ratio of dissolved strontium decreases with decreasing sampling elevation for the mainstem rivers (Figure 4.4). The samples from the Kosñipata river at Wayqecha and San Pedro have 87 Sr/ 86 Sr ratios that range from 80 0.727-0.728 and 0.719-0.721 respectively. All of the samples from the Alto Madre de Dios river at MLC have a similar 87 Sr/ 86 Sr ratio of 0.715 except for one sample collected in the wet season of 2013 that has a lower value of 0.708. Samples from the Madre de Dios river at CICRA all have a similar 87 Sr/ 86 Sr ratio of 0.709. None of the high Andean tributaries have 87 Sr/ 86 Sr ratios that are higher than the Kosñipata river at Wayqecha (Figure 4.4). However, the San Pedro river and R2050 tributaries have much lower 87 Sr/ 86 Sr ratios than the Kosñipata river (i.e. 0.708-0.717). Tributaries of the Alto Madre de Dios river at MLC also have low 87 Sr/ 86 Srratios(i.e. 0.708-0.709). Forelandfloodplaintributariesaremorevariable and have 87 Sr/ 86 Sr ratios that range from 0.709 to 0.718. 0 1 2 3 4 5 6 0.705 0.71 0.715 0.72 0.725 0.73 0.735 Ca/Na 87 Sr/ 86 Sr Way SP MLC CICRA mainstem tributary Figure 4.4: Radiogenic strontium isotope geochemistry. Purple symbols refer to samples from the Kosñipata River at Wayqecha catchment, red symbols refer to samples from the Kosñipata River at San Pedro catchment, Blue symbols refer to samples from Alto Madre de Dios River at MLC catchment, green symbols refer to samples from Madre de Dios River at CICRA catchment. Large filled diamonds are samples from a mainstem river and small open squares are samples from a tributary. Sulfate-sulfur isotope ratios Replicate purification of sulfate and measurement of its sulfur isotopic com- position from select samples reveals variability of up to 0.2h. This uncertainty 81 value is reported in all figures and tables but is not included after each value in the main text for brevity. Broadly, the δ 34 S value of dissolved SO 4 increases with decreasing sampling elevation for the mainstem rivers (Figure 4.5). The sulfur isotopic composition of dissolved SO 4 from the Kosñipata river at Wayqecha and San Pedro ranges from -2.1 to-0.5h and from-1.3 to-0.1h respectively. The sulfurisotopic composition of dissloved SO 4 from Alto Madre de Dios river at MLC ranges from -0.5 to +0.7 h. The sulfur isotopic composition of dissloved SO 4 from the Madre de Dios river at CICRA ranges from +2 to +5.5h. For the Madre de Dios river at CICRA, dry season δ 34 S values are higher than wet season values. Tributaries from the high Andes have highly variable δ 34 S values of dissolved SO 4 that range from -4.1 to +6.3h. These samples show a trend of increasing δ 34 S with decreasing SO 4 concentrations, though the samples from the San Pedro river, which have the lowest SO 4 concentrations, fall off of this trend. Tributaries from the low Andes and foreland floodplain catchments also have highly variable δ 34 S values (i.e. -3.8 to +6.9h), but display a positive relationship between δ 34 S values and SO 4 concentrations. Considering all of the tributaries together shows that samples with either high or low SO 4 concentrations typically have low δ 34 S values whereas samples with intermediate SO 4 concentrations are isotopically heavy. This bell-shaped relationship between SO 4 concentrations and isotopic ratios is observed in both the mainstem and tributary sample sets (Figure 4.5). Dissolved inorganic carbon isotope ratios Theδ 13 C values of samples collected during both the wet and dry seasons from the Kosñipata River at Wayqecha (n = 13) and San Pedro (n = 22) range -12.2 to -5.3h and -15.4 to -3.5h (Figure 4.6). For the Alto Madre de Dios river at MLC, measured δ 13 C values (n = 2) for the dry season are near -10h (Figure 4.6). For the Madre de Dios river at CICRA sampled during the dry season (n = 3), measured δ 13 C values range from -13.5 to -11.6 (Figure 4.6). WetanddryseasontributarysamplesfromtheKosñipataRiveratWayqecha(n = 10) show a similar carbon isotopic range to the mainstem samples (-20 to -7h; Figure 4.6). The samples with the highest δ 13 C values, which are from tributaries R2300, R2400, and R2432, stand out relative to the other samples in that they 82 10 0 10 1 10 2 10 3 −6 −4 −2 0 2 4 6 8 SO 4 μM δ 3 4 S δ 3 4 S Andean shales (n = 3) Way SP MLC CICRA mainstem tributary Figure 4.5: Sulfur isotope geochemistry. Purple symbols refer to samples from the Kosñipata River at Wayqecha catchment, red symbols refer to samples from the Kosñipata River at San Pedro catchment, Blue symbols refer to samples from Alto Madre de Dios River at MLC catchment, green symbols refer to samples from Madre de Dios River at CICRA catchment. Large filled diamonds are samples from a mainstem river and small open diamonds are samples from a tributary. The gray shaded region indicates the measured range in δ 34 S values of the solid phase samples have much lower DIC concentrations and define a positive linear trend with the inverse of the DIC concentration (Figure 4.6a). Wet and dry season tributary samples from the Kosñipata River at San Pedro (n = 10) show carbon isotopic ratios within the lower end of the range of the mainstem samples (-16.9 to -11.1 h; (Figure 4.6)). Wet season tributary samples from the Alto Madre de Dios River at MLC catchment show lower carbon isotope ratios than the mainstem during the dry season (-16.3 to -11.8h; (Figure 4.6)). Wet and dry season tributary samples from the Madre de Dios River at MLC catchment range from -18.9 to -11.6h (Figure 4.6). 83 It is worth noting that for many samples, the concentration of alkalinity cal- culated by charge balance is less than the DIC concentration measured by IS. A similar feature is seen in the Andes/Amazon dataset of Mayorga et al. (2005) for samples with circum-neutral pH values. Without pH and temperature measure- ments for the samples, this feature cannot be explored further. −30 −25 −20 −15 −10 −5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 100% 80% 60% 40% 100% 80% 60% 40% 20% H 2 SO 4 Silicate Weathering H 2 CO 3 Silicate Weathering or Respiration H 2 CO 3 Carbonate Weathering H 2 SO 4 Carbonate Weathering δ 13 C DIC (permil) SO 4 / (DIC + SO 4 ) 0 5 10 15 −30 −25 −20 −15 −10 −5 0 δ 13 C DIC (permil) 1000 / DIC A. B. Way SP MLC CICRA mainstem tributary Figure 4.6: Carbon isotope geochemistry. Purple symbols refer to samples from the Kosñipata River at Wayqecha catchment, red symbols refer to samples from the Kosñipata River at San Pedro catchment, Blue symbols refer to samples from Alto Madre de Dios River at MLC catchment, green symbols refer to samples from Madre de Dios River at CICRA catchment. Large filled squares are samples from a mainstem river and small open diamonds are samples from a tributary. (A) Variation inδ 13 C DIC values with the inverse of DIC concentrations. (B) Variation in δ 13 C DIC values with the SO 4 /(DIC+SO 4 ) ratio. The gray shaded region corre- sponds to mixing between sulfuric acid and carbonic acid weathering of carbonates for different stoichiometries and end-member compositions (see section 4.5.6). Val- ues along the edges of the gray shaded region refer to the percentage of sulfuric acid weathering of carbonates 4.4.3 Tributary mixing Combining all of the individual mixing fractions to estimate the contribution of each tributary to the total flux of the Madre de Dios river at CICRA requires 84 two important assumptions. First, it is assumed that discharges do not change significantly over the period of time that the samples were collected (i.e a couple of days). Second, it is assumed that the sampled tributaries account for a majority of the total discharge of the Madre de Dios river at CICRA. These assumptions can be tested by investigating the downstream evolution of δD for both the wet and dry season sampling cruises. If the assumptions are valid, then the isotopic composition of the mainstem river should only change at each confluence. Forboththewetanddryseasonsamplingcruises,somechangesintheδDvalues of the mainstem between confluences were observed (Figure 4.7). This suggests that the required assumptions were invalidated for some of the measurements. However, a sufficient amount of the data did meet the required assumptions and allows for a full water budget to be constructed for both time periods. Below, I first report the measurements that invalidated the assumptions. Subsequently, I present the median results of the water budget calculations. Additional percentiles of the results are reported in Table4.1. During the wet season sampling cruise, theδD values of the Alto Madre de Dios river at MLC are too low to match any of the downstream measurements taken during non-storm conditions. Similarly, the variability in the δD values of Alto Madre de Dios and Manu rivers observed between storm and non-storm conditions during the dry season brackets the values measured downstream. During both the wet and dry seasons, there is a small but unaccounted for increase in theδD value of the mainstem after its confluence with the Colorado River. For the wet season sampling cruise, the fact that the measurements of the Alto Madre de Dios river at MLC are too low does not directly affect any of the water budget calculations. For the dry season sampling cruise, I assume that the measurements downstream of the confluence of the Manu river reflect two extreme end-members (i.e. storm and non-storm) and that the true value is represented by the measurement of the Madre de Dios river upstream of the Chiribi river. For both sampling seasons, I assume that increase inδD observed after the confluence of the Colorado river is negligible, but then test this assumption by comparing the results to independent constraints (see below and 4.5.4). For the wet season sampling cruise, I calculate that the median contributions of the Alto Madre de Dios, Manu, Blanco, Chiribi, and Colorado rivers were 45%, 1%, 4%, 11%, and 33% of the total discharge of the Madre de Dios river measured 85 −65 −55 −45 −35 −25 −71.5 −71 −70.5 −70 −80 −78 −76 −74 −72 Longitude δD MLC MLC storm MdD us MdD ds-1 MdD ds-2 Manu Blanco Chiribi Colorado CICRA MLC MLC storm MdD us - s MdD ds - s MdD us MdD ds δD Dry Season Wet Season Upstream of conuence Downstream of conuence A B Figure4.7: Downstreamchangesinwaterisotoperatios. Squarescorrespond to samples collected downstream of a confluence. Circles correspond to samples collectedupstreamofaconfluence. Graybarshighlightwhenmultiplesampleswere collected due to the occurrence of storm events. The dashed lines schematically show how δD values evolve between confluences. (A) Downstream changes in δD during the dry season sampling campaign. (B) Downstream changes in δD during the wet season sampling campaign. at CICRA (Figure 4.8). For the dry season sampling cruise, I calculate that the median contributions of the Alto Madre de Dios, Manu, Blanco, Chiribi, and Colorado rivers were 10%, 34%, 1%, 6%, and 48% of the total discharge of the Madre de Dios river measured at CICRA (Figure 4.8). An independent estimate of the discharge of the Alto Madre de Dios river at the time of sampling can be generated by using the concentration-runoff relationships reported in Torres et al. (2015). For the dry season sampling cruise, the median concentration-runoff estimate (142 m 3 /sec) is in good agreement with the median value calculated using 86 the water isotope mixing approach (140 m 3 /sec). For the wet season sampling, the fact that the δD values of the Alto Madre de Dios river at MLC are too low to match any of the downstream measurements (see above) precludes the use of this sample as an independent estimate of water discharge. 0 200 400 600 800 1000 Alto Madre Manu Blanco Chiribi Colorado Dry Season Discharge (m 3 /sec) 0 1000 2000 3000 4000 Wet Season Discharge (m 3 /sec) A B Figure4.8: Waterdischargescalculatedbasedonwaterisotopemixing.(A) Box plots of calculated water discharges (m 3 /sec) during the wet season. (B) Box plots of calculated water discharges (m 3 /sec) during the dry season. 4.5 Discussion 4.5.1 Lithologic source partitioning - qualitative The observation that Na/Ca ratios remain constant while Mg/Ca ratios decrease from the Kosñipata river at Wayqecha to the Alto Madre de Dios river at MLC may be due to a shift from the weathering of Mg-rich to Mg-poor silicate rocks (Figure 4.3). This is consistent with the analyses of river bank sediments, 87 which suggest the presence of both a high and low Mg/Ca ratio silicate compo- nent that both have similar Na/Ca ratios. A downstream change in the relative contribution of different silicate sources to the dissolved load is further corrobo- rated by the coincident decrease in the dissolved 87 Sr/ 86 Sr ratio (Figure 4.4), which suggests a shift from the weathering of Mg-rich shales ( 87 Sr/ 86 Sr = 0.732±0.005; Dellinger et al. 2015) to more Mg-poor granitic rocks ( 87 Sr/ 86 Sr = 0.705; Dellinger et al. 2015). This would also explain the co-variation between Mg/Ca, Li/Ca, and SO 4 /Ca ratios, since shales are also enriched in Li-rich clays and sulfide minerals (Brantley et al., 2013; Dellinger et al., 2014). ThelowNa/CaandMg/CaratiosobservedfortheManuriverandtheMadrede DiosriveratCICRAareconsistentwithagreatercontributionfromthedissolution of carbonate minerals relative to the upstream sites (the Alto Madre de Dios river atMLCandtheKosñipatariver). Theriverbanksedimentleachates, whichshould reflect the chemistry of carbonate minerals, show low Mg/Ca, Na/Ca, and Li/Ca ratios as well as high Sr/Ca ratios that are consistent with the trends observed in the dissolved chemistry. Finally, the 87 Sr/ 86 Sr ratios of the the Manu river and the Madre de Dios river at CICRA, which range from 0.708-0.709, are consistent with a carbonate source (Andean carbonate 87 Sr/ 86 Sr = 0.7075±0.0005; Gaillardet et al. 1997). The high Mg/Ca and low Na/Ca ratios of tributaries of the Kosñipata river could be explained by the dissolution of dolomite. The Mg/Ca ratios of these rivers are close to 1, which is consistent with the stoichiometry of dolomite. Like Ca-carbonates, dolomites are also expected to have low Li/Ca ratios, which are observed in the Andean tributaries with high Mg/Ca and low Na/Ca ratios. Finally, some of the river bank sediment leachates showed Mg/Ca ratios as high as 0.6, which is consistent with the presence of some dolomite in addition to Ca- carbonates in the river bank sediments. The comparison between the major and trace elemental ratios of the dissolved and solid phase samples suggests that lithogenic solutes are sourced predominately from carbonate and silicate mineral dissolution. This is consistent with the obser- vation that more than 90% of the samples (i.e. 281/301) have Cl concentrations lower than 20μM, which is the maximum concentration of atmospherically sourced ClproposedbyGaillardetetal.(1997)fortheAndes/Amazonsystem. Thesamples that do exceed 20 μM Cl do not exceed 60 μM Cl. Rivers in the Andes/Amazon 88 system with known evaporite contributions (e.g., the Huallga River) typically have very high Cl concentrations (i.e. 848±421 μM Cl; Moquet et al. 2011). However, sulfate evaporites do not need to be systematically associated with chloride evaporites. Nevertheless, a significant contribution from sulfate evaporites to the dissolved load is unlikely due to the consistent and overall high Mg/Ca and Sr/Ca ratios observed in all of the sampled rivers. Sulfate salts such as gypsum (CaSO 4 ·2H 2 O) do not readily incorporate Mg or Sr resulting in very low Mg/Ca andSr/Caratios(<mmol/mol; Luetal.1997;Playà&Rosell2005). Whilesulfate evaporite deposits can also contain fluid inclusions and/or Mg and Sr sulfate salts, they are typically in insufficient quantities to increase the bulk ratios to the values seen for carbonate and silicate minerals (Lu et al., 1997; Playà & Rosell, 2005). This implies that sulfate in excess of what is supplied by precipitation is derived from the oxidation of sulfide minerals. This is discussed in further detail in section 4.5.3. 4.5.2 Lithologic source partitioning - quantitative To test my inferences about the dominant lithologic sources of solutes (section 4.5.1), I developed a two step mixing model that first uses a subset of the measured ratios (Cl/Ca, Na/Ca, Mg/Ca, and Sr/Ca) to calculate the proportions of solutes derived from each of the different sources by inversion and then forward models the remaining ratios (Li/Ca and 87 Sr/ 86 Sr) based on the calculated mixing pro- portions. If atmospheric deposition, two carbonate lithologies (Ca-carbonates and dolomites), and two silicate lithologies (shales and granitic rocks) are sufficient to explain the chemical composition of each of the sampled rivers, then the inversion componentofthemixingmodelshouldproducerealisticmixingproportionsthatin turn predict Li/Ca and 87 Sr/ 86 Sr ratios consistent with the actual measurements. Discrepancies between modeled and measured ratios imply either 1) missing solute sources, 2) missing solute sinks, 3) and/or a poor parameterization of the end- member ratios. Details of how I parameterize the model, its assumptions, and how I perform the calculations and uncertainty propagation are included in detail in Appendix F. For samples from the Kosñipata river catchments, the mixing model correctly predicts the measured Li/Ca ratios (Figure 4.9d), but significantly underestimates 89 the 87 Sr/ 86 Sr ratiofor mostsamples ifthe dolomiteend-member isassumed tohave the same Sr isotopic composition as Andean Ca-carbonates ((Figure 4.9a); Gail- lardet et al. 1997). The reason the model under-predicts the dissolved 87 Sr/ 86 Sr ratio for the Kosñipata river is that measured 87 Sr/ 86 Sr ratios are nearly as radio- genic as the shale end-member, which implies that most dissolved Sr, and by extension, Ca, is derived from silicate mineral dissolution (Figure 4.4). However, measured Na/Ca and Sr/Ca ratios are substantially lower than the shale mem- ber (Figure 4.3), which indicates a significant contribution from carbonates. The discrepancy between 87 Sr/ 86 Sr and elemental ratios is in the opposite sense as in the Himalayan system, where secondary carbonate precipitation causes elemental ratios to over-predict the contribution of silicate mineral dissolution relative to Sr isotope ratios (Jacobson et al., 2002; Bickle et al., 2015). The predicted and measured 87 Sr/ 86 Sr ratios can be made to agree if either the carbonate or silicate end-member is allowed to be more radiogenic than originally assumed (cf. Millot et al. 2003). For example, if the 87 Sr/ 86 Sr ratio of the dolomite end-memberisallowedtobeasradiogenicastheshaleend-member, thenthemodel can correctly re-produce both the measured Li/Ca and 87 Sr/ 86 Sr ratios for most of the samples (Figure 4.9b). Dolomite would be expected to have radiogenic Sr isotope compositions if it formed during the hydrothermal alteration of the shale (Evans et al., 2002). Alternatively, the model and data agree if the 87 Sr/ 86 Sr ratio of the shale end-member is allowed to have a 87 Sr/ 86 Sr ratio of∼0.9 (Figure 4.9c). This could possibly reflect the preferential dissolution of highly radiogenic sheet silicates within the shales (Blum & Erel, 1997). ForsamplesfromtheMadredeDiosrivercatchment,themixingmodelunderes- timates 87 Sr/ 86 SrratiosandoverestimatesmeasuredLi/Caratiosformanysamples if the dolomite end-member is assumed to have the same Sr isotopic composition as Andean Ca-carbonates (Figure 4.9a,d; Gaillardet et al. 1997). Like for the Kosñipata river catchments, the model can be forced to fit the strontium isotope data if the 87 Sr/ 86 Sr ratio of either the dolomite or shale end-member is increased (Figure 4.9b-c). For Li/Ca ratios, the ratio of the shale end-member would need to be decreased by an order of magnitude in order for the model to fit the data. Alternatively, there could be an unaccounted for sink of dissolved Li. The correla- tion between Li/Si and Li/Na ratios for samples from the Madre de Dios river at CICRA is consistent with a greater uptake of Li into secondary clays relative to 90 0.7 0.71 0.72 0.73 0.74 0.7 0.71 0.72 0.73 0.74 87 Sr / 86 Sr measured 87 Sr / 86 Sr modeled 0.7 0.71 0.72 0.73 0.74 0.7 0.71 0.72 0.73 0.74 87 Sr / 86 Sr measured 87 Sr / 86 Sr modeled 0.7 0.71 0.72 0.73 0.74 0.7 0.71 0.72 0.73 0.74 87 Sr / 86 Sr measured 87 Sr / 86 Sr modeled Limestone = 0.708 Dolomite = 0.708 Rain = 0.709 Granite = 0.705 Shale = 0.740 1:1 1:1 1:1 10 −4 10 −3 10 −2 10 −1 10 −4 10 −3 10 −2 10 −1 Li/Ca measured Li/Ca modeled 1:1 Dolomite = 0.740 Shale= 0.900 Way SP MLC CICRA mainstem tributary Figure 4.9: Lithologic mixing model. (A) A comparison of measured and modeled 87 Sr/ 86 Srvaluesforthereferencesetofend-memberratio(listedinpanel). (B) A comparison of measured and modeled 87 Sr/ 86 Sr values if the 87 Sr/ 86 Sr value of the dolomite end-member is as radiogenic as the shale end-member. (C) A comparison of measured and modeled 87 Sr/ 86 Sr values if the 87 Sr/ 86 Sr value of the limestoneend-memberisasradiogenicastheshaleend-member. (D)Acomparison of measured and modeled 87 Sr/ 86 Sr values if the 87 Sr/ 86 Sr value of the shale end- member is 0.9. A 1:1 line is shown on each panel for reference. the Kosñipata river catchment (Torres et al., 2015). Similarly, Li isotope studies point to floodplains as important loci of Li uptake into secondary clay minerals (Dellinger et al., 2015; Pogge von Strandmann & Henderson, 2014). I ascribe the misfit of the Li/Ca data in the Madre de Dios to such uptake (Figure 4.9). 4.5.3 Sulfate sources Since the oxidative dissolution of reduced sulfur minerals should not be accom- panied by significant isotopic fractionation (Balci et al., 2007), the measured δ 34 S 91 of reduced sulfur phases in rock samples can be used to predict the sulfur isotopic composition of rivers where the SO 4 budget is dominated by sulfide mineral oxida- tion (Figure 4.5). The overlap between the isotopic composition of the Kosñipata river at Wayqecha, its tributaries, and the measured δ 34 S of reduced sulfur phases in rock samples suggests that the sulfate budget of these rivers is dominated by sulfide mineral oxidation. The samples of the mainstem rivers and tributaries that have δ 34 S values of dissolved SO 4 above what is seen in the rock samples (i.e. >0h). This suggests that either 1) the rock samples do not encompass the entire range of variability in the δ 34 S of sulfide minerals, 2) there is an additional sulfate source, or 3) there is a secondary process that fractionates sulfur isotope ratios. While it may be likely that the three rock samples do not capture the entire range of variability in δ 34 S, the correlation between SO 4 concentrations and δ 34 S is unexpected based on the measurements of δ 34 S and wt.% pyrite in the rock samples. One known additional source of sulfate is from atmospheric deposition. Rain- waters often have SO 4 /Cl ratios greater than seawater, which suggests that SO 4 is added to precipitation as a result of biological or anthropogenic processes (Andreae etal.,1990a,b). Consequently,thesulfurisotopiccompositionofatmosphericdepo- sition should be lighter than modern seawater (i.e. < 21h; Paris et al. 2013). Two measurements of Amazonian rainwater taken from Brazil and Columbia have δ 34 S values of +10 to +11h, confirming the importance of biogenic sulfur aerosols to the isotopic composition of precipitation (Longinelli & Edmond, 1983). However, since these samples were taken much further north than the study site, they cannot be used as a robust constraint on the sulfur isotopic composition of atmospheric SO 4 . Instead, samples of pure foreland floodplain tributaries, like the Los Amigos and Chirbi rivers, may be better constraints since they are likely to derive most of their SO 4 from atmospheric deposition. These tributaries show very low SO 4 concentrations (i.e. ≤ 10 μM) and low δ 34 S values (i.e. < 0h). This suggests that the highδ 34 S values observed in some of the rivers are not due to a significant atmospheric contribution. This conclusion is consistent with measured Cl concen- trations and local rainwater SO 4 /Cl ratios, which imply that only a small fraction of measured sulfate is derived from atmospheric deposition (Torres et al., 2015). 92 Sulfate evaporite minerals are another potential source of sulfate and would be expected to have high δ 34 S values. However, as discussed above in section 4.5.2, elemental and isotopic budgets are inconsistent with a significant evaporite contribution. For example, theδ 34 S values of the Manu river, which are the highest of all the rivers measured, would imply that more than 50% of its sulfate was derived from Paleozoic evaporites (δ 34 S = +10; Longinelli & Edmond 1983). This levelofevaporiteinputshouldcausemeasurabledecreasesintheMg/CaandSr/Ca ratios of the Manu river that are not observed. Sulfate reduction is a common process that fractionates sulfur isotopes. Biota preferentially reduce 32 S leaving the residual sulfate pool enriched in 34 S (Detmers et al., 2001). Reduction of dissolved SO 4 is consistent with the observed trend of decreasing SO 4 concentrations with increasing δ 34 S values for the high to inter- mediate SO 4 concentration samples. However, in order for sulfate reduction to have an effect on the sulfur isotopic composition of the Kosñipata river and its tributaries, the product, H 2 S, cannot be re-oxidized within the catchment if the system is at or near geomorphic steady-state (i.e. uplift≈ denudation). To explain a similar trend in the δ 34 S of sulfate in Himalayan rivers, Turchyn et al. (2013) proposed that some H 2 S produced by sulfate reduction is re-precipitated as a sec- ondary sulfide mineral. For this mechanism to work under steady-state conditions, it would require that secondary sulfide minerals are exported from the catchment without being re-oxidized. The mass of secondary sulfide minerals that needs to be exported from the catchment in order to maintain a steady-state isotopic offset between dissolved sulfate and primary sulfide minerals is dependent upon the isotopic fractionation associated with sulfate reduction. Using the formulation of Bouchez et al. (2013), the isotopic offset of the river relative to primary sulfide minerals (Δδ 34 S Diss−Prim ) can be expressed as: Δδ 34 S Diss−Prim =−Δ sec−sulf ×F (4.1) where Δ sec−sulf is the mean isotopic offset between dissolved sulfate and secondary sulfidemineralsandFisthefractionofthetotalSfluxthatisexportedasareduced solid phase. In natural systems, Δ sec−sulf is variable, but typically on the order of -10h. Thus, for the δ 34 S of riverine sulfate to be increased by 1h, F needs to 93 be on the order of 0.1. The trade-offs between F and Δ sec−sulf to drive changes in Δδ 34 S Diss−Prim implied by equation 4.1 are shown graphically in Figure 4.10. Based on the sulfate concentration-runoff relationship of Torres et al. (2015) and the runoff time-series of Clark et al. (2014), the annual flux of dissolved sul- fate for the Kosñipata river at San Pedro is on the order of 10 3 tons/yr. A similar approach, reported in Clark et al. (2014b), yields a suspended sediment flux on the order of 10 5 tons of sediment/yr. Thus, for the isotopic composition of the Kosñi- pata river at San Pedro to be offset by 1h with a Δ sulfide value of -10h, these order of magnitude estimates of fluxes require suspended sediments to contain, on average, 0.1 wt. % S. This is a testable prediction that should be considered in future work on sulfur cycling in this and other mountainous catchment systems. 4.5.4 Tributary mixing The tributary mixing model reveals that the relative contribution of each trib- utary to the total water flux of the Madre de Dios river measured at CICRA is seasonally variable (Figure 4.8). Thus, conservative mixing alone should lead to significant seasonal variations in solute concentrations and elemental ratios. In addition to variation in the mixing proportions of different tributaries, non- conservative behavior of solutes can also drive variation in the chemical compo- sition of the Madre de Dios river measured at CICRA. The relative importance of conservative mixing versus non-conservative behavior for different solutes can be diagnosed by comparing the measured elemental ratios at CICRA to modeled ratios based on the mixing fractions calculated using the downstream change in the water isotope composition of the Madre de Dios river (see section 4.4.3). For the wet season sampling cruise, the measured and modeled Na/Cl and SO 4 /Cl ratios of the Madre de Dios river at CICRA overlap, which is consistent with conservative mixing (Figure 4.11a). However, measured Mg/Cl, Ca/Cl, Sr/Cl ratios are all higher than would be predicted based on conservative mixing (Figure 4.11b-d). Similarly, the measured Li/Cl ratio is lower than would be predicted based on conservative mixing (Figure 4.11c). For the dry season sampling cruise, the measured Na/Cl ratio is slightly lower than the modeled results (Figure 4.12a). Since it is unlikely that Na/Cl ratios would be affected by non-conservative processes, this offset is likely due to some 94 0 0.1 0.2 0.3 0.4 10 9 8 7 6 5 4 3 2 1 0 Fraction of S exported as a reduced phase Δ δ 34 S Rock−River Δ = -5 Δ = -10 Δ = -15 Δ = -20 Δ = -25 Figure 4.10: S isotope fractionation at geomorphic steady state. Variations in the maximum isotopic offset of dissolved sulfate relative to a rock source at geo- morphic steady state (i.e. uplift≈ denudation) for different values of the isotopic fractionation factor (Δ sec−sulf ) and the fraction of S exported as a reduced phase calculated using the model of Bouchez et al. (2013). of the simplifications and assumptions that were made during the construction of the water budget (see section 4.4.3). The model can be forced to fit the measured Na/ClratiobydecreasingthecontributionofeithertheManuorColoradoriversby approximately 0.2 or 0.3 respectively (Figure 4.12a). Decreasing the contribution of the Manu river allows the model to match both the Na/Cl and SO 4 /Cl ratios (Figure4.12a). Importantly,thereisnowaytoforcethemodeltosimultaneouslyfit all of the measured elemental ratios. Instead, like observed during the wet season, measured Mg/Cl, Ca/Cl, Sr/Cl ratios are all too high where as the measured Li/Cl ratio is too low (Figure 4.12b-d). 95 20 30 40 50 0 50 100 150 200 250 Na/Cl Ca/Cl 20 30 40 50 0 5 10 15 20 25 30 Na/Cl SO 4 /Cl 20 30 40 50 0 10 20 30 40 50 60 Na/Cl Mg/Cl 20 30 40 50 0.04 0.05 0.06 0.07 0.08 0.09 0.1 Na/Cl Li/Cl Manu Alto Madre Colorado Blanco Chiribi CICRA D C A B Modeled Wet Season Figure 4.11: Modeled elemental ratios based on conservative tributary mixing. Modeled elemental ratios of the Madre de Dios at CICRA (gray points) based on conservative mixing between the Manu (red), Alto Madre (cyan), Blanco (green), Chiribi (blue), and Colorado (magenta) Rivers during wet season condi- tions using the mixing fractions based on conservative tracers (Figure 4.8; section 4.4.3). Measured elemental ratio of the Madre de Dios River at CICRA are show for comparison (yellow star). (A) Na/Cl and SO 4 /Cl ratios. (B) Na/Cl and Mg/Cl ratios. (C) Na/Cl and Li/Cl ratios. (D) Na/Cl and Ca/Cl ratios. The observation that measured Ca/Cl, Mg/Cl, and Sr/Cl are all excess of what would be predicted based on conservative mixing implies that Ca, Mg, and Sr are added to dissolved load by an additional process (Figures 4.11-4.12). One likely additional source of these elements is the dissolution of carbonate minerals present as suspended sediments. The selective loss of Ca from suspended sediments in the adjacent Beni river foreland floodplain has previously been attributed to carbonate 96 dissolution by Bouchez et al. (2012), which supports suggestion that a similar process may operate in the Madre de Dios river foreland floodplain. The observation that the measured Li/Cl ratios are lower than what would be predicted based on conservative mixing implies that Li released during weathering intheAndes isremoved from solutionduringfloodplain transit(Figures4.11-4.12). Since Li readily partitions into clay minerals (Huh et al., 2004b), the precipitation ofthesephasesmayexplainthelossofLifromsolution. InarecentLiisotopestudy of Amazonian rivers, Dellinger et al. (2015) observed large Li isotope fraction dur- ing floodplain transit for rivers with headwaters in the Andes. Since Li is strongly fractionated during uptake into secondary clays, the observations of Dellinger et al. (2015) are consistent with my observation of non-conservative behavior of Li. For both the dry and wet season sampling cruises, the measured and mod- eled SO 4 /Cl ratios are in close agreement. This implies that SO 4 behaves nearly conservatively during floodplain transit. This result is interesting since the sulfur isotopic composition of sulfate from the Madre de Dios river measured at CICRA varies seasonally by at least 3.5h. Owing to their high SO 4 concentrations, both the Alto Madre de Dios and Manu rivers dominate the SO 4 budget of the Madre de Dios river at CICRA. Additionally, these two rivers display very distinct δ 34 S values. Thus, seasonal changes in the mixing proportions of these two rivers may explain much of the observed sulfur isotopic variation. The water isotope mixing calculations reveal that during the dry season, the Manu river contributes more to the total flux of water and sulfate of the Madre de Dios river at CICRA than the Alto Madre de Dios river. This pattern is consistent withtheobservationthatduringthedryseason, theδ 34 SoftheMadredeDiosriver at CICRA is elevated (i.e. +4.3 to +5.5h) and closer to the value of the Manu river (+6.4h). During the wet season, the Alto Madre de Dios river contributes more to the total flux of water and sulfate than the Manu river. Again, this is consistent with the observation that theδ 34 S of the Madre de Dios river at CICRA is lower (i.e. +2 to +2.8h) during the wet season and is closer to the value of the Alto Madre de Dios river, which ranges from -1 to 0h. 97 4.5.5 The acid budget of chemical weathering The results of the lithologic mixing model (section 4.5.2) and the sulfur isotope measurements (section 4.5.3) can be used together to construct an acid budget of chemical weathering for each of the sampled rivers. For this calculation, I use the calculated fractions of Ca, Mg, Na, and K sourced from silicate mineral dissolu- tion, their uncertainties (1σ), and measured solute concentrations to estimate the sum of silicate cations (in units of alkalinity equivalents) for each sample and its uncertainty. Since the mixing model did not consider K, I use the approach of Torres et al. (2015) to partition K between silicate and atmospheric sources. Note that the calculated silicate alkalinity is not necessarily equal to the actual silicate alkalinity because some sulfuric acid may drive silicate weathering. The calculated silicate alkalinity is then compared to the sulfate acidity, which is calculated as the difference between the total sulfate concentration and the amount that can be attributed to atmospheric deposition determined using the approach of Torres et al. (2015). This partitioning of sulfate is based on the results of the lithologic mixing model (section 4.5.2) and the sulfur isotope mea- surements (section 4.5.3), which suggest that evaporite minerals do not contribute significantly to the sulfate budget of this system. For the Kosñipata River at Wayqecha and San Pedro, calculated silicate alka- linity is similar in magnitude to sulfate acidity (Figure 4.13). This results suggests that chemical weathering in this river catchment, which drains a rapidly eroding portion of the Andes, does not contribute to long-term CO 2 consumption due to the effects of sulfide mineral oxidation. The Madre de Dios River at CICRA has both lower calculated silicate alkalinity and lower sulfate acidity relative to the the Kosñipata River (Figure 4.13). However, for the Madre de Dios at CICRA, calcu- lated silicate alkalinity is significantly in excess of sulfate acidity, which suggests that chemical weathering in this river catchment, which drains the slowly eroding foreland floodplain, does contribute to long-term CO 2 consumption (Figure 4.13). Importantly, the general results of the acid budget calculation are not sensitive to the assumption that evaporites do not contribute significantly to the sulfur budget in this system. All of the samples from Kosñipata River catchment at Wayqecha, which show the largest amount of CO 2 release (Figure 4.13), have δ 34 S values that overlap with the measured pyrite samples and are thus consistent 98 with pyrite being the dominant sulfate source (Figure 4.5). Samples from the foreland floodplain catchments, which show net CO 2 consumption (Figure 4.13), have δ 34 S values that could possibly reflect sulfate evaporite dissolution (Figure 4.5). In this case, sulfate evaporite dissolution would act to increase the amount of CO 2 consumption in the foreland floodplain and amplify the difference in the acid budgets between rapidly eroding (Kosñipata River; CO 2 release) and depositional (Madre de Dios River; CO 2 consumption) settings (Figure 4.13. 4.5.6 Riverine DIC Based on a mixing model of major and trace element ratios (section 4.5.2) and S isotopic measurements (section 4.5.3), I infer that the proportion of chemical weathering driven by sulfuric acid decreases from the Andes Mountains to the foreland floodplain (section 4.5.5). As an additional test of this hypothesis, I utilize δ 13 C DIC as a tracer of DIC sources and cycling. In soils, rivers, and ground waters without hydrothermal inputs, there exist three major sources of DIC: organic matter respiration, carbonate minerals, and atmospheric CO 2 (Sackett & Moore, 1966; Wigley et al., 1978; Schulte et al., 2011; Galy & France-Lanord, 1999; Li et al., 2008). Since these sources typically have distinct δ 13 C values (Sackett & Moore, 1966; Wigley et al., 1978; Schulte et al., 2011; Galy & France-Lanord, 1999; Li et al., 2008), changes in the relative contributions from each source should result in variations in the δ 13 C of DIC. The reaction of carbonate minerals with carbonic acid sourced from organic matter respiration or sulfuric acid sourced from sulfide oxidation following the equations CaCO 3 +H 2 CO 3 ↔Ca 2+ + 2HCO − 3 (4.2) and CaCO 3 +H 2 SO 4 ↔Ca 2+ +H 2 CO 3 +SO 2− 4 (4.3) should produce DIC with different δ 13 C values since DIC is sourced solely from carbonate minerals during sulfuric acid weathering (equation 4.3; Galy & France- Lanord 1999; Li et al. 2008. However, variations in DIC sources are not the sole control on the δ 13 C values of DIC in rivers. 99 In addition to variations in DIC sources (e.g., equations 4.2-4.3), the carbon isotopic composition of DIC in rivers can also be affected by CO 2 degassing, equili- bration with atmospheric or soil CO 2 , carbonate precipitation, and/or the uptake of CO 2 by photosynthetic organisms (Sackett & Moore, 1966; Wigley et al., 1978; Doctor & Kendall, 2008; Li et al., 2008; Schulte et al., 2011; Polsenaere & Gwenael, 2012). With the exception of equilibration with soil CO 2 , these processes tend to increase the δ 13 C value of the residual pool of riverine DIC relative to its initial value (Sackett & Moore, 1966; Wigley et al., 1978; Doctor & Kendall, 2008; Li et al., 2008; Schulte et al., 2011; Polsenaere & Gwenael, 2012). Ultimately, riverine DIC should approach isotopic equilibrium with respect to the atmosphere (Sackett & Moore, 1966; Schulte et al., 2011; Polsenaere & Gwenael, 2012). In the following discussion, I will mainly consider the effects of CO 2 degassing and equilibration with atmospheric CO 2 since soil storage is thought to be a small component of the overall hydrologic budget in this system (Clark et al., 2014) and the the studied rivers are turbid (i.e. in-situ photosynthesis is likely minimal; Dokulil 1994) and undersaturated with respect to carbonates. Organic matter respiration and/or the sulfuric acid weathering of carbonate minerals can supersaturate river waters with CO 2 relative to the atmosphere and drive CO 2 degassing (Sackett & Moore, 1966; Doctor & Kendall, 2008; Butman & Raymond, 2011; Schulte et al., 2011; Polsenaere & Gwenael, 2012). As discussed in Polsenaere & Gwenael (2012), CO 2 degassing from rivers with pH values between 4.7 and 7 affects only dissolved CO 2 concentrations and not alkalinity. However, boththe equilibrium isotopicoffsetsbetweenthedifferentDICspecies(i.e. carbon- ate, bicarbonate, and dissolved CO 2 ) and kinetic processes can result in significant isotopic fractionation during CO 2 degassing (Wigley et al., 1978; Zhang & J, 1995; Polsenaere & Gwenael, 2012). Equilibration with the atmosphere should drive the δ 13 C value of DIC towards 0 to +1h (Schulte et al., 2011; Polsenaere & Gwe- nael, 2012). Unfortunately, the effects of degassing and equilibration are difficult to distinguish from variations in DIC sources. Since many samples have DIC con- centrations in excess of alkalinity calculated by charge balance and the carbon isotopic composition of many of the rivers without "excess" DIC are lighter than would be predicted for isotopic equilibrium with the atmosphere, it is possible that many of the studied rivers are not significantly impacted by degassing and isotopic equilibration. 100 Following the approach of Galy & France-Lanord (1999), which assumes that the effects of degassing and equilibration are negligible, I compare δ 13 C DIC val- ues to the ratio of SO 4 concentrations to the sum of SO 4 and DIC concentrations ( SO 4 (DIC+SO 4 ) ; Figure 4.6b). The measurements are then compared to mixing rela- tionships based on reaction stoichiometries (e.g., equations 4.2 - 4.3), measured δ 13 C values of detrital carbonates (section 4.4.1), and the organic carbonδ 13 C val- ues presented in Clark et al. (2013). In Figure 4.6, I show the trajectories of how silicate weathering by either carbonic or sulfuric acid would affect the theoretical mixing relationships, but do not explicitly model silicate weathering processes. Most samples show δ 13 C DIC and SO 4 (DIC+SO 4 ) ratios consistent with a combina- tion of sulfuric and carbonic acid weathering of carbonate minerals (Figure 4.6b). Notably, samples of the Kosñipata River at Wayqecha and its tributaries have higher SO 4 (DIC+SO 4 ) ratios than would be predicted by 4.2-4.3 Figure 4.6b. Based on the linear relationship between δ 13 C DIC and the inverse of DIC concentrations (Figure 4.6a), I hypothesize that the some of the tributary samples are affected by CO 2 degassing. This may also explain the offset of the mainstem samples, though consideration of the major element budgets suggests that these samples likely also reflect some contribution of silicate weathering by sulfuric acid. Overall, theδ 13 C DIC measurements are qualitatively consistent with the results of the mixing model (section 4.5.2) and S isotopic measurements (section 4.5.3). Andean samples have higher δ 13 C DIC values that are consistent with carbonate weathering by sulfuric acid (Figure 4.6). The lowerδ 13 C DIC values of the foreland floodplain samples are consistent with carbonate weathering by carbonic acid. Thus, despite the uncertainties and assumptions that underlie each of the geo- chemical approaches, the overall conclusion of a geomorphic control on the acid budget of chemical weathering is robust. 4.6 Conclusions Identification of this geomorphic control on the acid budget of chemical weath- ering has important implications for the effects of mountain uplift on the long-term carbon, oxygen, andsulfurcycles(Torresetal.,2014). Likeinotherrapidlyeroding mountainous systems (e.g., Galy & France-Lanord 1999; Calmels et al. 2007, 2011; 101 Das et al. 2012; Torres et al. 2014), chemical weathering in the Andean Kosñipata river catchment does not contribute significantly to long-term CO 2 consumption and may instead result in CO 2 release into the ocean/atmosphere system due to the effects of sulfuric acid (Figure 4.13). On the other hand, chemical weathering in the foreland floodplain may be an important CO 2 sink (Figure 4.13) despite being less sensitive to seasonal changes in runoff (Torres et al., 2015). Another important observation is that the flux of sulfate from sulfide oxidation fromtheKosñipatariveratSanPedroissimilarinmagnitudetothefluxofmodern particulate organic carbon (Clark et al., 2014). This suggests that O 2 consumption by oxidation reactions is in excess of O 2 production by organic carbon export for Kosñipata river catchment. This difference in O 2 consumption versus production may be further exacerbated by the oxidation of fossil organic carbon (Bouchez et al., 2010; Hilton et al., 2014). Since sulfide oxidation reactions are likely limited to mountainous catchments, downstream changes in the river loading of modern particulate organic carbon that occur in the foreland floodplain may be key to the net effect of surface processes on the fluxes of O 2 . River Catchment area (km 2 ) Season Median discharge (m 3 /sec 95% CI lower bound 95% CI upper bound MLC 6025 dry 141 114 169 wet 1959 953 3355 Manu 13479 dry 497 407 574 wet 54 0 529 Blanco 982 dry 15 0 97 wet 156 0 870 Chiribi 1145 dry 83 30 136 wet 516 0 1544 Colorado 3648 dry 699 627 780 wet 1453 127 2446 Table 4.1: Calculated water discharges based on conservative tracers 102 10 15 20 25 0 20 40 60 80 100 Na/Cl Ca/Cl 10 15 20 25 0 5 10 15 20 Na/Cl SO 4 /Cl 10 15 20 25 4 6 8 10 12 14 16 18 20 22 Na/Cl Mg/Cl 10 15 20 25 0 0.01 0.02 0.03 0.04 0.05 Na/Cl Li/Cl Manu Alto Madre Colorado Blanco Chiribi CICRA Modeled Modeled (Fraction Manu - 0.2) D C A B Dry Season Figure 4.12: Modeled elemental ratios based on conservative tributary mixing. Modeled elemental ratios of the Madre de Dios at CICRA (gray points) based on conservative mixing between the Manu (red), Alto Madre (cyan), Blanco (green), Chiribi (blue), and Colorado (magenta) Rivers during dry season condi- tions using the mixing fractions based on conservative tracers (Figure 4.8; section 4.4.3). Measured elemental ratio of the Madre de Dios River at CICRA are show for comparison (yellow star). The pink points indicate modeled elemental ratios calculated after reducing the contribution of the Manu River by 0.2.(A) Na/Cl and SO 4 /Cl ratios. (B) Na/Cl and Mg/Cl ratios. (C) Na/Cl and Li/Cl ratios. (D) Na/Cl and Ca/Cl ratios. 103 1 10 100 1000 10 100 1000 Sulfate Acidity (µM) Silicate Alkalinity (µEq) CO 2 Consumption Way SP MLC CICRA 1:1 CO 2 Release 2:1 Figure 4.13: The acid budget of weathering in the Kosñipata / Madre de Dios River system. A comparison between the alkalinity equivalent of the sum of silicate cation concentrations calculated using the mixing model (section 4.5.2) with the amount of atmospheric-input corrected sulfate concentration calculated using the approach of Torres et al. (2015). Values to the right of the 1:1 line are equivalent to CO 2 release during weathering. Values to the left of the 1:1 line are equivalent to CO 2 consumption during weathering. This approach assumes that rainwater and pyrite are the sole SO 4 sources. However, as described in section 4.5.5, this assumption does not affect the broad trends in CO 2 release versus CO 2 consumption. Purple symbols refer to samples from the Kosñipata River at Wayqecha catchment, red symbols refer to samples from the Kosñipata River at San Pedro catchment, Blue symbols refer to samples from Alto Madre de Dios River at MLC catchment, green symbols refer to samples from Madre de Dios River at CICRA catchment. 104 Chapter 5 Sulphide oxidation and carbonate dissolution as a source of CO 2 over geological timescales 5.1 Linking Statement The results of Chapter 4 suggest that tectonic uplift may drive the release of CO 2 into the ocean/atmosphere system by promoting the oxidation of trace sulfide minerals. If generally applicable, this result has important implications for the effects of tectonic activity on the long-term cycling of CO 2 and O 2 . In this chapter, the implications of tectonically-enhanced sulfide oxidation are explored using the Cenozoic period as a test case. Dr. Joshua West is thanked for suggesting that I pursue this project beyond the single slide that I made for a departmental seminar. Dr. Gaojun Li is thanked for incorporating sulfide oxidation into his inversion model (Li et al., 2009; Li & Elderfield, 2013) and generating Figure 5.2. A manuscript based on this work was published in Nature (Torres et al., 2014). 5.2 Introduction and Main Text The observed long-term (million-year) stability of Earth’s climate is thought to depend on the rate of CO 2 release from the solid Earth being balanced by the rate of CO 2 consumption by silicate weathering (Walker et al., 1981). During the Cenozoic period spanning the last approximately 65 Ma, the concurrent rise in the marine isotopic ratios of Sr, Os, and Li (McArthur et al., 2001; Klemm et al., 2005; Misra & Froelich, 2012) suggests that extensive uplift of mountain ranges may have stimulated CO 2 consumption by silicate weathering (Raymo et al., 1988), but reconstructions of seafloor spreading (Müller et al., 2008) do not 105 indicate a corresponding increase in CO 2 inputs from volcanic degassing. The resulting imbalance would have depleted the atmosphere of all CO 2 within a few million years (Berner & Caldeira, 1997). As such, reconciling Cenozoic isotopic records with the need for mass balance in the long-term carbon cycle has been a major and as yet unresolved challenge in geochemistry and Earth history. Here I show that enhanced sulfide oxidation coupled to carbonate dissolution can provide a transient source of CO 2 to Earth’s atmosphere that is significant over geolog- ical timescales. Like drawdown via silicate weathering, this source is probably enhanced by tectonic uplift and so may have contributed, at least in part, to the relative stability of Cenozoic atmospheric pCO 2 . A variety of other hypotheses (e.g., Raymo & Ruddiman 1992; Bickle 1996; Li & Elderfield 2013) have been put forward in order to explain the "Cenozoic isotope-weathering paradox", and the evolution of the carbon cycle likely depended on multiple processes. However, a significant role for sulfide oxidation coupled to carbonate dissolution is consistent with records of radiogenic isotopes (McArthur et al., 2001; Klemm et al., 2005), atmospheric pCO 2 (Pagani et al., 2005; Zhang et al., 2013), and the evolution of theCenozoicsulfurcycle, andcouldbeaccommodatedwithgeologicallyreasonable changes in the global O 2 cycle, suggesting that this CO 2 source should be consid- ered as a potentially important but as-yet generally unrecognized component of the long-term carbon cycle. IthasbeenproposedthatthatexcessCO 2 uptakefromenhancedsilicateweath- ering over the Cenozoic could be balanced by a net decrease in the size of the organic C reservoir (Raymo & Ruddiman, 1992). However, recent work has shown that orogenic activity generally increases organic C burial and stabilizes petrogenic organic C, through graphitization (Galy et al., 2008). Model results also suggest that global changes in organic matter burial during the Cenozoic were insufficient to balance excess CO 2 consumption (e.g., Katz et al. 2005; Li & Elderfield 2013). The release of CO 2 from the thermal decomposition of carbonate minerals dur- ing metamorphism has been suggested to be the missing source of CO 2 required to balance the Cenozoic C cycle (Bickle, 1996). Petrologic and geochronologic data on the timing and potential magnitude of associated fluxes call this mecha- nism into question (Kerrick & Caldeira, 1998). Additionally, any estimate of CO 2 release during metamorphism is subject to considerable uncertainty because it is not known what fraction of the CO 2 makes it into the atmosphere after passing 106 through kilometers of variably permeable and reactive rock (Kerrick & Caldeira, 1998). Recent modeling work by Li & Elderfield (2013) has suggested that the acceler- ation of continental weathering by Cenozoic uplift was compensated by a factor of approximately 2 decrease in the weathering of ocean island basalts in response to changing climatic conditions. This is not implausible but requires very significant changes, and there is currently no direct evidence to test this prediction. Another mechanism that has not yet been systematically explored is that Ceno- zoic uplift, in addition to increasing rates of CO 2 drawdown by silicate weathering, increased rates of sulfide oxidation coupled to carbonate dissolution, which pro- vided a transient source of CO 2 . The oxidation of sulfides (e.g. pyrite) produces sulfuric acid following the equation: 4FeS 2 + 15O 2 + 14H 2 O↔ 4Fe(OH) 3 + 8H 2 SO 4 (5.1) The dissolution of carbonate minerals with this sulfuric acid results in the trans- fer of sedimentary C into the ocean/atmosphere system as CO 2 , which can be represented by the equations: CaCO 3 +H 2 SO 4 ↔CO 2(g) +H 2 O +Ca 2+ +SO 2− 4 (5.2) and 2CaCO 3 +H 2 SO 4 ↔ 2Ca 2+ +SO 2− 4 + 2HCO − 3 (5.3) For equation 5.2, the release of CO 2 occurs immediately. For equation 5.3, CO 2 is still released but on the timescale of carbonate precipitation in the ocean (10 6 yrs; Berner & Berner 2012), represented by the equation: Ca 2+ + 2HCO − 3 ↔CaCO 3 +CO 2(g) +H 2 O (5.4) A strong correlation between erosion and sulfide oxidation rates (Calmels et al., 2007) suggests that orogenic activity, in addition to increasing CO 2 drawdown through enhanced silicate weathering, will also increase CO 2 release from coupled sulfide oxidation and carbonate dissolution. 107 Estimates of the modern sulfide oxidation fluxes range from 0.64 to 1.26 Tmol yr −1 and are based on either the chemistry of large rivers (Berner & Berner, 2012) or the modeled composition of weathering sources (Lerman et al., 2007). Direct measurements of sulfide derived S fluxes have been limited, as isotopic measure- ments are generally required in order to reliably distinguish between sulfate and sulfide-derived S. Data from the localities where the relative proportions of sulfide and sulfate derived S have been constrained indicate a flux of approximately 0.2 Tmol/yr of sulfide S, contributed by approximately 2% of global land area (Galy & France-Lanord, 1999; Millot et al., 2003; Calmels et al., 2007, 2011; Das et al., 2012). Directly extrapolating this value globally would yield a flux of 10 Tmol/yr sulfide S. While the areas with well-constrained data may represent environments withdisproportionatelyhighsulfideoxidationsrates, itisalsolikelythattheactual global sulfide oxidation rate has been previously underestimated, or at least lies in the upper range of previous estimates. In the localities where both silicate weathering and sulfide oxidation rates have been measured, it is possible to compare the rates of CO 2 uptake and release (Figure5.1; seesection5.3). FortheMackenzieRiverinCanada(Millotetal.,2003; Calmels et al., 2007) and the Liwu River in Taiwan (Calmels et al., 2011; Das et al., 2012), the calculated release rates of CO 2 associated with sulfide oxidation are greater than the uptake rates of CO 2 by silicate weathering, suggesting that these environments release CO 2 into the atmosphere. For the Ganges-Brahmaputra in India (Galy & France-Lanord, 1999), the rates of CO 2 consumption and release are comparable, whichsuggeststhattheentireCO 2 consumptionbysilicateweathering is, to first order, balanced by coupled sulfide oxidation and carbonate dissolution. Ultimately, the CO 2 released by coupled sulfide oxidation and carbonate dis- solution is balanced by alkalinity production during sulfate reduction and sulfide precipitation in marine sediments, which can be represented by the generalized equation: 2FeOOH+2H + +Ca 2+ +2SO 2− 4 +4CH 2 O↔FeS 2 +Fe 2+ +4H 2 O+Ca 2+ +4HCO − 3 (5.5) The long residence time of sulfate (>10 7 ) relative to Ca 2+ (10 6 ) in the ocean allows forthetransientreleaseofsedimentaryCintotheatmosphereasaresultofcoupled 108 10 4 10 5 10 6 10 7 Mackenzie G-B Liwu mol CO 2 yr -1 km -2 RELEASE CONSUMPTION Figure 5.1: CO 2 uptake (green bars) and release (red bars) rates calcu- lated for the Mackenzie River (Millot et al., 2003; Calmels et al., 2007), the Liwu River in Taiwan (Calmels et al., 2011; Das et al., 2012), and the Ganges-Bhramaputra (G-B; Galy & France-Lanord 1999). The com- piled data are presented in Table 5.1 and a description of the calculation method is presented in section 5.3. Error bars reflect the propagation of the analytical uncer- tainty of each value that was reported by the original authors (1 s.d.). For the Ganges-Bhramaputra, the uncertainty in the proportion of the sulfate flux derived from sulfide oxidation was not reported and therefore no error bar is included. sulfide oxidation and carbonate dissolution (Calmels et al., 2007; Berner & Berner, 2012). The controls on the duration of transient CO 2 release can be explored using a conceptual box model of the ocean (Figure 5.2; see Methods). In this model, the input of dissolved inorganic carbon (DIC) derived from carbonate weathering by sulfuric acid (equations 5.1,5.2, and 5.3) provides excess sedimentary DIC to the ocean, reflecting CO 2 release. The excess sedimentary DIC is ultimately balanced by DIC derived from the oxidation of organic carbon during sulfate reduction (equation 5.5) such that the ratio of sedimentary DIC to total DIC derived from carbonateweatheringisequalto0.5atsteadystate. Foraninstantaneousdoubling of weathering-derived input fluxes, an increase in the ratio of sedimentary DIC to 109 total DIC is observed for a period of 70 Myr reflecting a transient release of CO 2 (Figure 5.2). For a doubling of input fluxes that occurs over a period of 40 Myr, the transient release of CO 2 lasts for 95 Myr (Figure 5.2). These calculated timescales are comparable to the duration of enhanced silicate weathering implied by isotopic records, suggesting that enhanced sulfide oxidation has the potential to play an important role in countering the excess CO 2 drawdown caused by tectonic uplift during the Cenozoic. Sulfide minerals are concentrated in organic-rich sedimentary rocks along with chalcophile elements such as Re and Os (Georg et al., 2013). The decay of 187 Re to 187 Os leads to elevated 187 Os/ 188 Os ratios in sulfide/organic rich sedimentary rocks, makingthemarinerecordof 187 Os/ 188 Ossensitivetotheweatheringofthese lithologiesovergeologictime(Georgetal.,2013). Usingthemodelingframeworkof Li & Elderfield (2013), which reconstructs changes in sedimentary rock weathering based in part on a 187 Os/ 188 Os mass balance, I calculate the CO 2 release as a result of coupled sulfide oxidation and carbonate dissolution during the Cenozoic (see section 5.3). The effects of pyrite burial are not included in the model as there are few constraints on how burial rates may have varied over the Cenozoic. Using the model output, I calculate the change in the net CO 2 drawdown from mountain uplift throughout the Cenozoic (Figure 5.2). The reconstructed net CO 2 drawdown from mountain uplift is relatively constant from 50-25 Ma and consis- tent with proxy-based pCO 2 reconstructions (Figure 5.2; Pagani et al. 2005; Zhang et al. 2013). From 25-15 Ma there is a period of marked increase in the net CO 2 consumption from mountain uplift that corresponds with a significant decline in pCO 2 (Figure 5.2; Pagani et al. 2005; Zhang et al. 2013). This increase results from a mismatch in the shape of the 87 Sr/ 86 Sr and 187 Os/ 188 Os curves such that silicate weathering continues to increase while the weathering of sulfide/organic rich sedimentary rocks remains constant (Figure 5.2). After this perturbation, the net CO 2 drawdown from mountain uplift and the pCO 2 of the atmosphere both stabilize at respectively higher and lower values relative to 50-25 Ma (Figure 5.2, Pagani et al. 2005; Zhang et al. 2013). Lower pCO 2 conditions from 15-0 Ma would decrease weathering fluxes due to the climate-weathering feedback and act to balance the larger net CO 2 drawdown from mountain uplift over this time period (e.g., Li & Elderfield 2013), which is consistent with 10 Be/ 9 Be evidence for 110 constant global weathering fluxes over the last 12 Ma (Willenbring & von Blanck- enburg, 2010). The general consistency between the proxy-based pCO 2 record and the model calculations supports the hypothesis that sulfide oxidation played an important role in in the Cenozoic C cycle and may help, at least in part, to recon- cile radiogenic isotope records with atmospheric pCO 2 proxies and the requirement of mass balance in the long-term carbon cycle. Enhanced sulfide oxidation in response to Cenozoic uplift should also impact the long-term sulfur cycle. Fluid inclusion analyses (Horita et al., 2002) and the isotopic composition of authigenic barites (Paytan et al., 1998) suggest that marine sulfate concentrations increased significantly during the Cenozoic without a cor- responding change in the sulfur isotope ratio. This complicates straightforward interpretation of the Cenozoic S cycle as processes that could modulate marine sulfate concentrations should also affect sulfur isotope ratios. By using an inverse model of the S cycle (see section 5.3), I find that my reconstruction of sedimentary rock weathering can be reconciled with the marine δ 34 S sulfate record for a rea- sonable range of parameters values (Figures 5.5-5.7) and that the corresponding range of CO 2 release is of similar magnitude to what would be required to balance hypothesized increases in silicate weathering (Figure 5.8). Thus, my hypothesis is consistent with the marine δ 34 S sulfate record at the same time that it can explain the observed increase in sulfate concentrations through the Cenozoic. The O 2 content of the atmosphere depends on the balance between the burial andoxidationofreducedcarbonandsulfurcompounds(Berner,2000). Anincrease in sulfide oxidation will consume O 2 , with a molar ratio of O 2 consumption to CO 2 production of 15/8 (see equations equations 5.1,5.2, and 5.3; Figure 5.4). The hypothesis that CO 2 release as a result of sulfide oxidation was important in bal- ancing enhanced CO 2 consumption since the Miocene (as suggested by Figure 5.2) would require O 2 consumption during this period. Completely compensating for a 25%increaseinsilicateweatheringCO 2 drawdownoverthelast15Ma(theincrease implied in the model of Li & Elderfield (2013) and shown in Figure 5.2 would mean an early-mid Miocene pO2 of approximately 25%, compared to present day 21% (Figure 5.4). There is little independent evidence for how much atmospheric pO 2 varied over Cenozoic timescales, but changes on the order of approximately 4% are within geologically reasonable ranges (Watson et al., 1978; Berner, 2000; Wildman et al., 2004) and consistent with estimates derived from the sulfur cycle modeling 111 (see section 5.3; Figure 5.8). If O 2 consumption occurred over the whole Cenozoic, as some of the S isotope models suggest (see section 5.3; Figure 5.5), the addi- tional O 2 demand could require Paleocene pO 2 values on the order of 23-33% (see section 5.3). Importantly, these calculations do not fully capture actual expected changes in pO 2 , because the oxygen budget also depends on fluxes in the organic carbon cycle (Berner, 2000) and the average oxidation state of this carbon, which is variable (LaRowe & Van Cappellen, 2011) and poorly constrained. The key point is that the O 2 -CO 2 trade-off places a practical limit on geologically reasonable CO 2 production by sulfide oxidation, but the relative magnitudes of the reservoirs and fluxes mean that sulfide oxidation can play a significant role in balancing sil- icate weathering-driven CO 2 consumption if accompanied by modest variation in atmospheric pO 2 . It is unlikely that the evolution of the carbon cycle during the Cenozoic depended alone on changes in silicate weathering balanced by sulfide-carbonate weathering; additional CO 2 release from metamorphic decarbonation (Bickle, 1996), decreasing rates of ocean island basalt weathering associated with global climate-weathering feedbacks (Li & Elderfield, 2013), and changes in the organic carbon cycle may also have been important. Although many questions remain, including about the coincident variation in pO 2 , sulfide oxidation coupled to car- bonate dissolution does provide an attractive and thus far under-recognized source of CO 2 because it is driven by the same forcing mechanism as enhanced CO 2 con- sumption by silicate weathering (Calmels et al., 2007), operates on a sufficiently long time scale (Figure 5.2), and is consistent with proxy records of weathering fluxes (McArthur et al., 2001; Klemm et al., 2005; Misra & Froelich, 2012) and atmospheric pCO 2 over Cenozoic timescales (Figure 5.2; Pagani et al. 2005; Zhang et al. 2013). 5.3 Methods 5.3.1 Calculation of CO 2 Fluxes from Modern Rivers All data for calculating the river fluxes shown in Figure 5.1 have been compiled from published literature. CO 2 uptake rates are calculated from fluxes of Na, K, Ca, and Mg corrected for rainwater input using Cl − concentrations and rainwater 112 cation/Cl − ratios. To partition the Ca flux between carbonate and silicate com- ponents, end-member silicate Ca/Na ratio of 0.35 was used for the Liwu (Calmels et al., 2011) and Mackenzie (Millot et al., 2003) while a value of 0.2 was used for the Ganges-Brahmaputra (G-B; Galy & France-Lanord 1999). To partition the Mg flux between silicate and carbonate sources, an end-member silicate Mg/Na ratios of 0.24 was used for the Liwu (Calmels et al., 2011) and Mackenzie (Millot et al., 2003) whereas a Mg/K ratio of 0.5 was used for the G-B (Galy & France- Lanord, 1999) following the methods of the original authors. All of the Na and K fluxes after correction for rain contributions are assumed to be derived from silicate sources except in the Mackenzie, where a portion of Na flux is contributed by evaporites and was accounted for using 87 Sr/ 86 Sr ratios assuming an evaporite end-member ratio of 0.7081 (Millot et al., 2003). All of the silicate Ca and Mg fluxes were assumed to contribute to long-term CO 2 drawdown whereas only 30% of the Na flux and 20% of the K flux were considered, following FranceâĂŘLanord & Derry (1997). For the G-B calculations, some of the values differ slightly from those reported in FranceâĂŘLanord & Derry (1997), but this does not affect the overall interpretation. The calculated values are presented in Table 5.1. To calculate CO 2 release rates, the proportion of the sulfate flux derived from sulfide oxidation was estimated using δ 34 S sulfate (Liwu, Das et al. 2012), δ 34 S sulfate and δ 18 O sulfate (Mackenzie, Calmels et al. 2007), or δ 13 C DIC (Ganges- Brahmaputra, Galy & France-Lanord 1999). All of the sulfide-derived SO 2− 4 flux was assumed to react preferentially with carbonates, which provides a maximum estimate of CO 2 release rates. The relative difference between release and uptake rates (i.e., the comparison shown in Figure 5.1) does not change depending upon the partitioning of the sulfide-derived SO 2− 4 between silicate and carbonate rocks. For the Liwu calculation, the δ 34 S sulfate measurements used to calculate the pro- portion of sulfide-derived SO 2− 4 are from the adjacent Kaoping River (Das et al., 2012) as there are no published values from the Liwu River. Annual fluxes from the Liwu are considered because they are better constrained, and the two rivers drain similar lithologies (Calmels et al., 2007; Das et al., 2012). 113 mol/yr/km 2 Na + silicate K + silicate Mg 2+ silicate Ca 2+ silicate CO 2 consumption (uncertainty) Liwu 2.6E+05 1.7E+05 6.3E+04 9.3E+04 2.1E+05 (6.7E+04) GB 6.8E+04 2.8E+04 1.4E+04 1.4E+04 4.0E+04 (1.6E+04) Mackenzie 2.1E+04 3.5E+03 5.3E+03 6.8E+03 1.6E+04 (4.7E+03) Ca 2+ carbonate Mg 2+ carbonate SO 2− 4 total SO 2− 4 sulfide CO 2 release (uncertainty) Liwu 3.2E+06 9.1E+05 2.1E+06 1.8E+06 1.8E+06 (1.3E+05) GB 2.3E+05 8.9E+04 7.9E+04 5.5E+04 5.5E+04 Mackenzie 1.4E+05 6.4E+04 7.6E+04 6.2E+04 6.2E+04 (3.1E+03) Table 5.1: Large river CO 2 fluxes 5.3.2 Conceptual Box Model The conceptual box model (Figure 5.2) was constructed using a set of ordinary differential equations describing the concentrations of DIC and sulfate in the ocean considering only the fluxes related to the weathering/precipitation of carbonates and sulfides (i.e. equations 5.1, 5.2, 5.3, 5.4, and 5.5). Initial values of 8×10 18 moles/Myr, 5×10 18 moles/Myr, and 1×10 18 moles/Myr are used for the input fluxes of total DIC, sedimentary DIC, and sulfide derived sulfate respectively. The output fluxes (carbonate precipitation and sulfate reduction/sulfide precipitation) are assumed to vary linearly as a function of concentration with rate constants of 1 Myr −1 for DIC and 0.1 Myr −1 for sulfate (Berner & Berner, 2012). The model was run for 120 Myr in order to reach steady state before the input fluxes were doubled instantaneously (Figure 5.2a) or over the course of 40 Myr (Figure 5.2b). 5.3.3 Li and Elderfield Mass Balance Model The results from the adapted Li & Elderfield (2013) model shown in Figure 5.2 are based on a series of isotope mass balance equations that relate the weath- ering of continental silicates, the weathering of sedimentary rocks (which includes C org and carbonate weathering), the weathering of basaltic rocks, reverse weath- ering/hydrothermal alteration, C org burial, carbonate burial, and volcanic CO 2 degassing to the marine records of 87 Sr/ 86 Sr (McArthur et al., 2001), 187 Os/ 188 Os (Klemm et al., 2005), δ 13 C org (Falkowski et al., 2005), δ 13 C carbonate (Falkowski et al., 2005), dissolved Mg concentrations (Horita et al., 2002), and seafloor spreading rates (Müller et al., 2008). The mass balance equations are solved simultaneously using smoothed fits of the published marine proxy records in order to calculate changes in the fluxes of the different C cycle parameters relative to modern day 114 values (the modern values used in the calculation are presented in Li & Elderfield 2013). To simplify the model, various assumptions are made regarding the mass bal- ances of each element/isotope system. For the Os system, the isotopic composition of the ocean is considered to be controlled by the weighted average of inputs from continental silicate rocks, island basalts, sedimentary rocks, hydrothermal activity, and cosmic dust as described by the equation: Σ[K i ×Os i × (R Os i −R Os ocean )] = 0 (5.6) where K is the intensity of the ith flux relative to the late Pleistocene value, Os i is the late Pleistocene value of the ith flux, R Os is the normalized osmium isotopic composition of the ith flux, and R Os ocean is the normalized osmium isotopic composi- tion of the ocean (Li & Elderfield, 2013). This is a significant simplification of the Os system because different sedimentary lithologies contribute different amounts of Os with distinct 187 Os/ 188 Os values. For example, studies of modern river sys- tems (Levasseur et al., 1999; Huh et al., 2004a) have suggested that carbonates contribute unradiogenic Os whereas organic-rich shales contribute more radiogenic Os. This is taken into account in the model by assigning a mean 187 Os/ 188 Os value for sedimentary rocks that reflects a weighted contribution from different sedimen- tary lithologies (Li & Elderfield, 2013). Since this value is held constant over the Cenozoic, I assume that weathering fluxes from different sedimentary lithologies change in a constant proportion. This is a reasonable assumption because differ- ent sedimentary lithologies are inter-bedded at a large scale, and thus exhumed in tandem. If this proportion were to vary, one would predict that Os fluxes from car- bonates would increase relative to organic-rich shales because carbonate minerals are more abundant (Lerman et al., 2007). This would result in an underestimation of changes in sedimentary rock weathering making the calculation of CO 2 release from sulfide oxidation a minimum estimate, thereby strengthening the argument about the potential importance of this mechanism. The Os system is further sim- plified by assuming a constant cosmic input flux. 115 Due to the long residence time of Sr relative to the modeling timescale, the transient-state solution to the mass balance equation is used and expressed as: N× dR Sr ocean dt = Σ[K i ×Sr i × (R Sr i −R Sr ocean )] (5.7) where N is the amount of Sr in seawater, dR Sr /dt is the time derivative of normal- ized Sr isotopic composition of the ocean, Sr i is the late Pleistocene value of the ith flux, R Sr is the normalized Sr isotopic composition of the ith flux, and R Sr ocean is the normalized R Sr ocean isotopic composition of the ocean (Li & Elderfield, 2013). In this mass balance, contributions from continental silicate rocks, island basalts, sedimentary rocks (i.e. carbonates), hydrothermal activity, and sediment diagen- esis are considered. Like the Os isotope system, the reconstruction of weathering fluxes using the marine 87 Sr/ 86 Sr record is sensitive to the isotopic composition of source rocks. While carbonate minerals typically have low 87 Sr/ 86 Sr values rela- tive to igneous rocks, intense metamorphic activity, such as that experienced in the Himalayan region, can impart unusually radiogenic Sr isotope compositions to carbonate minerals (Edmond, 1992). The potential for radiogenic Sr fluxes from metamorphic carbonates in the Himalayan region to bias the reconstruction of silicate weathering fluxes was previously assessed by Li et al. (2009) and Li & Elderfield (2013). This sensitivity analysis was conducted by modeling a linear increase in the isotopic composition of the Sr flux from Tibetan rivers (from 0.7077 to 0.7081) over the last 40 Ma and partitioning this flux in variable proportions between 100% carbonate and 100% silicate sources in order to test the effects of increasing input of radiogenic Sr from Tibetan rivers over the Cenozoic on the reconstruction of weathering fluxes. Consideration of these end-member scenarios minimally affected the reconstructed silicate weathering fluxes (Li et al., 2009; Li & Elderfield, 2013). To incorporate CO 2 release due to coupled sulfide oxidation and carbonate dissolutionintothismodel, modernratesofsulfideoxidationweretakentobeequal to 50% of the modern natural riverine S flux (Berner & Berner, 2012) and kept in a constant proportion to the total weathering flux of sedimentary rocks, which is calculated based on the mass balances of 87 Sr/ 86 Sr (equation 5.7), 187 Os/ 188 Os (equation 5.6), δ 13 C, and alkalinity, but is most sensitive to the 187 Os/ 188 Os mass balance. The net weathering CO 2 drawdown associated with uplift (as shown in 116 Figure 5.2) is calculated by taking the difference between the CO 2 consumption by silicate weathering and CO 2 release from coupled carbonate dissolution and sulfide oxidation. This approach assumes that a majority of sulfides are hosted in sedimentary rocks as it neglects contributions from igneous sulfide minerals. If igneous sulfide oxidation in a globally important process, the model provides an upperestimateofthenetweatheringCO 2 drawdownassociatedwithupliftbecause it underestimates the CO 2 flux from sulfide oxidation. 5.3.4 Sulfur Isotope Mass Balance Model In order to determine how the pyrite burial flux, the δ 34 S of riverine input, and the isotopic fractionation factor associated with sulfate reduction and pyrite burial (Δ pyrite−seawater ) would have to vary in order to reconcile the marine δ 34 S record with the reconstruction of sedimentary rock weathering from the (Li & Elderfield, 2013) model, the non-steady-state inverse modeling approach of (Kurtz et al., 2003) was used. In this approach, the differential equation describing the isotopic mass balance of S, dδ 34 S dt = F riverine × (δ 34 S riverine −δ 34 S ocean )−F pyriteburial × Δ pyrite−seawater Ms (5.8) where δ 34 S/dt is the time rate of change of the isotopic composition of marine sulfate, F riverine is the input flux of S from rivers (10 18 mol Ma −1 ), δ 34 S riverine is the isotopic composition of the riverine input,δ 34 S ocean is the isotopic composition of marine sulfate, F pyriteburial is the burial flux of S as pyrite (10 18 mol Ma −1 ), Δ pyrite−seawater is the average fractionation factor associated with sulfate reduction and pyrite burial, and Ms is the mass of marine sulfate (10 18 mol), is rearranged to calculate the time evolution of the parameter of interest. To constrain the time variation in F riverine , I use my reconstruction of sedimentary rock weathering from the(Li&Elderfield,2013)massbalancemodel(Figure5.2)assumingthatthetotal S yield (i.e. evaporite + sulfide) from sedimentary rock weathering has remained constantandthatthemoderndayriverinesulfatefluxisequalto3×10 18 molS/Ma (Berner & Berner, 2012). By modeling the riverine input flux in this manner, I assume that both evaporite and sulfide oxidation fluxes increased in concert over the Cenozoic in response to tectonic uplift, which is geologically reasonable as 117 different sedimentary lithologies are interbedded at a large scale. As a result, the isotopic composition of the riverine input does not have to vary over the Cenozoic in the model. Nevertheless, the potential for variation in the isotopic composition of the riverine input to influence the Cenozoic sulfur cycle was explored using the inverse model calculations (see below). To constrain the time variation in the mass of marine sulfate, I use an expo- nential fit to the fluid inclusion data from Horita et al. (2002) following Kurtz et al. (2003). To constrain the time dependent term dδ 34 S/dt, I calculate the first derivative of a smoothing spline fit of the δ 34 S barite record (Paytan et al., 1998) after a 20-point robust loess smoothing of the raw data. This fit to the δ 34 S barite record is also used for the value of δ 34 S ocean in the model. The time variation of the fractionation factor associated with sulfate reduction and pyrite burial (Δ pyrite−seawater ), the S isotopic composition of riverine input (δ 34 S riverine ), and the pyrite burial flux (F pyrite−burial ) are all poorly constrained. As a result, it is necessary to perform separate model experiments where two of the parameters values are fixed at constant values and the third is calculated. In this study, three types of model experiments were conducted, (1) variable pyrite burial fluxes with constant δ 34 S riverine and Δ pyrite−seawater , (2) variable δ 34 S riverine with constant F pyrite−burial and Δ pyrite−seawater and (3) variable Δ pyrite−seawater with constant F pyrite−burial and δ 34 S riverine . To calculate the implied changes in the pyrite burial flux (model type 1), Δ pyrite−seawater was varied between -50 and - 30h following Wu et al. (2010) and the δ 34 S riverine input was varied between 0 and +10h following Refs. Paytan et al. (1998) and Kurtz et al. (2003). It is important to note that this range of δ 34 S values for the riverine input does not necessarily reflect the balance between evaporite and sulfide sources as each source displayssignificantisotopicvariability(Wuetal.,2010)andriverinesulfatemaybe significantly fractionated by biological processes prior to being discharged into the ocean (Turchyn et al., 2013). To calculate the implied changes in Δ pyrite−seawater (model type 2), the δ 34 S of riverine input was varied between 0 and +10h and the pyrite burial flux was kept at a constant value of either 0.22, 0.67, or 1.0×10 18 mol S Ma −1 following Paytan et al. (1998) and Kurtz et al. (2003). To calculate the implied changes in theδ 34 S of riverine input (model type 3), Δ pyrite−seawater was variedbetween-50and-30handthepyriteburialfluxwaskeptataconstantvalue of either 0.22, 0.67, or 1.0×10 18 mol S Ma −1 . The constant pyrite burial fluxes 118 greaterthan0.22×10 18 molSMa −1 usedinthemodelcalculationsof Δ pyrite−seawater and δ 34 S riverine can be greater than the riverine input flux reconstructed from the Li & Elderfield (2013) isotope mass balance for a few time periods between 30-50 Ma. For this reason, the model experiments for variableδ 34 S of riverine input and Δ pyrite−seawater were only evaluated between 0-30 Ma. 5.4 Supplementary Results 5.4.1 Sulfur Cycle Mass Balance Model Results from solving the sulfur mass balance model for variable pyrite burial flux For the model experiments where the mass balance equation was arranged to solve for the pyrite burial flux with constant values for the δ 34 S of riverine input and Δ pyrite−seawater , the calculated pyrite burial fluxes were directly proportional to the riverine input flux. The constants of proportionality varied depending upon the δ 34 S of riverine input and Δ pyrite−seawater . Lower values of the δ 34 S of riverine input and higher values of Δ pyrite−seawater produced higher pyrite burial fluxes for the same riverine input flux (Figure 5.5). Multiple model experiments were consistent with the long-term average burial flux estimated by Holser et al. (1988) from the pyrite content of marine sediments (Figure 5.5). For the period around 45-50 Ma, where there is a major positive excursion in the marine δ 34 S record26, some of the model experiments produced pyrite burial fluxes greater than the total riverine input flux (Figure 5.5). After this brief period, all of the model experiments produced burial fluxes that were lower than the total riverine input flux (Figure 5.5). For model experiments where the δ 34 S of riverine input was equal to 10h, calculated pyrite burial fluxes were less than the estimated sulfide oxidation fluxes used in the Li & Elderfield (2013) mass balance model (i.e. 50% of the total S flux) regardless of the value of Δ pyrite−seawater (Figure 5.5). For model experiments where the δ 34 S of riverine input was equal to 5h, calculated pyrite burialfluxeswerelessthantheestimatedsulfideoxidationfluxesfor Δ pyrite−seawater values <-30h (Figure 5.5). 119 For model experiments where the δ 34 S of riverine input was equal to 0h, cal- culated pyrite burial fluxes were less than the estimated sulfide oxidation fluxes for Δ pyrite−seawater values <-40h (Figure 5.5). For reference, the Phanerozoic (0-200 Ma) value of Δ pyrite−seawater estimated from the average isotopic offset between coeval pyrite and marine sulfate is -43± 2h (Wu et al., 2010). Results from solving the sulfur mass balance model for variable isotopic fractionation during pyrite burial For the model experiments where the mass balance equation was arranged to solve for Δ pyrite−seawater with constant values for the pyrite burial flux and the δ 34 S of riverine input, higher δ 34 S values of riverine input and lower burial fluxes resulted in lower Δ pyrite−seawater values for the same input flux (Figure 5.6). High burial fluxes and/or high values of δ 34 S riverine were required in order to maintain Δ pyrite−seawater values above -60h, which has been suggested to be the maximum extent of isotopic fractionation during the Phanerozoic (Wu et al., 2010; Habicht et al., 2002). All model experiments produced values that were distinct from estimates based on the measured average isotopic offset between coeval pyrite and marine sulfate for at least a portion of the period from 0-30 Ma39. From 0-12 Ma, each model experiment showed decreasing Δ pyrite−seawater values (Figure 5.6). Results from solving the sulfur mass balance model for variable isotopic composition from rivers For the model experiments where the mass balance equation was arranged to solvefortheδ 34 Sofriverineinputwithconstantvaluesforthepyriteburialfluxand Δ pyrite−seawater , larger values of Δ pyrite−seawater and higher burial fluxes resulted in lowerδ 34 S values of riverine input for the same input flux (Figure 5.7). From 0-12 Ma, model experiments where the burial flux was≥ 0.67×10 18 mol S Ma −1 showed increasing values of theδ 34 S of riverine input (Figure 5.7). These experiments also produced calculated δ 34 S values of riverine input at 0 Ma near estimates for the modern value(Paytan et al., 1998; Kurtz et al., 2003). 120 Role of increased input fluxes of sulfur to the oceans IncreasingCenozoicmarinesulfateconcentrationsrequirevariationintheinput and/or output fluxes of sulfate. Since the variation of a single parameter within the S cycle would result in a change in the isotopic composition of marine sulfate, co-variation of different parameters during the Cenozoic is necessary in order to reconcile the relative constancy of the marine δ 34 S record (Paytan et al., 1998) with increasing marine sulfate concentrations (Horita et al., 2002). The results presented here suggest that for a reasonable range of parameter values, increasing marine sulfate concentrations during the Cenozoic could result from increasing input fluxes from weathering if pyrite burial rates, Δ pyrite−seawater , and/or the δ 34 S of riverine input co-vary with the input flux (Figures 5.5-5.7). Despite this agreement for some of the model experiments, a mechanism for the co-variation with the riverine input flux is required. The factors that control the fractionation factor associated with sulfate reduc- tion and pyrite burial are incompletely understood, but Δ pyrite−seawater is thought to be positively correlated with sulfate concentrations (i.e. higher [SO 2− 4 ]; Habicht et al. 2002) leads to less isotopic fractionation). This behavior is inconsistent with my model as a negative correlation between sulfate concentrations and Δ pyrite−seawater is necessary in order for the δ 34 S record to be reconciled with increasing input fluxes by only changing Δ pyrite−seawater (Figure 5.6). Addition- ally, the model calculations of Δ pyrite−seawater are not in good agreement with independent estimates from the sedimentary record (Wu et al., 2010). As a result, it is unlikely that variation in Δ pyrite−seawater alone can explain variations in the Cenozoic S cycle. Model experiments where the burial flux of S was≥ 0.67×10 18 mol S Ma −1 produced δ 34 S riverine values in agreement with independent estimates (Paytan et al., 1998; Kurtz et al., 2003). In these experiments, the relative constancy of the δ 34 S of marine sulfate was partly maintained in spite of increasing input fluxes through a significant increase in the δ 34 S value of riverine input from 0-12 Ma (Figure 5.7). This increase could arise from a change in the source of sulfate in response to tectonic uplift. An increase in the portion of sulfate derived from evaporite weathering from < 50% during a majority of the Cenozoic to the modern value of approximately 50% starting at approximately 12 Ma would sufficiently 121 increaseintheδ 34 Svalueofriverineinput. Suchanincreaseinevaporiteweathering would mean that I have underestimated the effects of sulfide oxidation in the isotope mass balance model (Figure 5.2), because I assume that the portion of riverine sulfate derived from sulfide oxidation remained at the modern value for the entirety of the Cenozoic when it may have been greater during the early Cenozoic. It is also possible to increase the δ 34 S value of riverine input without changing the proportion of sulfide-derived sulfate, as sulfide minerals themselves display a wide range of isotopic values depending upon the depositional environment and the geologicage(Wuetal.,2010). Similarly, agreaterfluxofsulfatefromenvironments wherethemicrobialprocessingofriverinesulfateingroundwatersissignificant(e.g. the Himalayan system; Turchyn et al. 2013) could produce an increase in the δ 34 S value of riverine input without changing the proportion derived from sulfide oxida- tion. Increasing pyrite burial fluxes could also reconcile the Cenozoic marine δ 34 S curve with enhanced riverine fluxes and increasing sulfate concentrations (Figure 5.5). This linkage is consistent with the known coupling between organic carbon and pyrite burial fluxes (Berner, 1982) and the importance of physical erosion in driving enhanced rates of sulfide oxidation and organic carbon burial (Calmels et al., 2007; Galy et al., 2007). However, a tight coupling between riverine input and pyrite burial fluxes could potentially limit the timescale of transient CO 2 release in response to sulfide oxidation. The sensitivity analysis employed in this study sug- gests that for a reasonable range ofδ 34 S values of riverine input and Δ pyrite−seawater (see above), sulfide oxidation fluxes can remain in excess of pyrite burial fluxes over the Cenozoic, which is required for transient CO 2 release (Figure 5.5). While the magnitude of CO 2 release from sulfide oxidation will depend upon the difference between pyrite oxidation and burial fluxes, the overall increase in CO 2 release dur- ing the Cenozoic is preserved. This maintains the general trend of changes in the net weathering CO 2 sink from tectonic uplift throughout the Cenozoic calculated with the (Li & Elderfield, 2013) mass balance model (Figure 5.2). Role of decreased pyrite burial fluxes Besidesanincreaseintheriverineinputflux, increasingCenozoicmarinesulfate concentrations could also be explained by a long-term decrease in pyrite burial fluxes. Unlike increasing riverine fluxes in response to tectonic uplift, it is unclear 122 what processes could lead to a decrease pyrite burial fluxes and not affect the isotopic composition of marine sulfate. However, it is important to note that decreasing pyrite burial fluxes over the Cenozoic would lead to an imbalance in the sulfur cycle and CO 2 release into the ocean/atmosphere system in the same manner as enhanced continental sulfide oxidation. Summary of the sulfur cycle modeling The plausible variation in the δ 34 S of riverine input and pyrite burial fluxes determined for a reasonable range or parameter values by the model calculations supports the idea that sulfide oxidation was enhanced during Cenozoic in response totectonicuplift. Inadditiontoprovidingameanstoreconcilemyhypothesiswith the marine δ 34 S record, these results also help to explain the observed increase in marine sulfate concentrations during the Cenozoic (Horita et al., 2002). Regardless of the specific driver, the modeled scenarios produce conditions under which CO 2 is released into the ocean/atmosphere system in response to sulfide oxidation fluxes in excess of sulfide burial fluxes. 5.4.2 Determining the effect of sulfide oxidation on Ceno- zoic pO 2 To assess the trade offs between O 2 consumption and CO 2 production implied by my model, I considered a molar ratio of O 2 consumption to CO 2 production for pure sulfide oxidation of 15/8, which is based on equation 5.1 and is consistent with previousworkquantifyingtheeffectofsulfurredoxtransformationsonatmospheric oxygen (Berner, 2000). I have hypothesized that sulfide oxidation may have played a significant role in balancing enhanced CO 2 consumption by silicate weathering over the last 15 Ma (e.g., Figure 5.2), so I have considered the specific case of the O 2 -CO 2 trade-off over this time interval (Figure 5.4). I converted the O 2 consumption values to the required pO 2 at 15 Ma by using a value of 37.8×10 18 moles for the present day mass of O 2 in ocean/atmosphere system and treating the atmosphere as an ideal gas (i.e. mole fraction = volume fraction). To compare these pO 2 consumption values to the CO 2 fluxes, CO 2 drawdown from silicate weathering was calculated for linear increases of 10%, 25%, and 50% from 15 Ma to 123 modern day values. The range from 10-50% increase brackets the range of previous predictions (Bickle, 1996; Li & Elderfield, 2013) and includes the values implied by Figure 5.2. A modern day silicate weathering flux of 3.5×10 18 mols C Ma −1 was calculated from the flux data of Meybeck & Ragu (1996) and Gaillardet et al. (1999) and the global lithologic partitioning of Gaillardet et al. (1999) assuming that100%oftheriverinesilicateCa 2+ andMg 2+ fluxcontributedtoCO 2 drawdown while 30% of the Na + flux and 20% of the K + flux contributed after exchanging for Ca 2+ in marine sediments (following the approach of FranceâĂŘLanord & Derry 1997, and as described above). The implied pO 2 at 15 Ma, as considered in this analysis, can be readily compared to the fire limit of 25% O 2 suggested by (Watson et al., 1978), although the experiments of (Wildman et al., 2004) suggest that the fire limit is above 35% O 2 , and the modeling results of (Berner, 2000) suggest a Phanerozoic maximum value of 35% O 2 . To determine the O 2 consumption implied by the S cycle inverse models, the net sulfur flux was calculated by subtracting the sulfide burial flux from the sulfide oxidation flux at million year intervals. The net sulfide flux was integrated over the period from either 0-50 Ma or 0-15 Ma after performing a cubic interpolation using MATLAB. The integrated value was converted from moles of sulfur to moles O 2 consumption using a ratio of O 2 consumption to sulfide oxidation of 15/8 (equa- tion 5.1). Implied Paleocene and Miocene pO 2 values were determined, as above, based on a present-day O 2 inventory in the atmosphere of 37.8×10 18 moles, and by treating the atmosphere as an ideal gas (i.e. mole fraction = volume fraction). Implications of enhanced sulfide oxidation for Cenozoic pO 2 Integratedoverthelast50Ma, theO 2 consumptionimpliedbytheresultsofthe variable sulfide burial models varies from -26.8×10 18 to 27.5×10 18 moles O 2 where negative values reflect net O 2 production (Figure 5.8a). Net oxygen production is a consequence of calculated pyrite burial fluxes that are higher than the sulfide oxidation fluxes used as model input (see above). Integrated over the last 15 Ma, O 2 consumption varies from -9.9×10 18 to 14.5×10 18 moles O 2 (Figure 5.8b). For all of the model experiments where pyrite burial fluxes were fixed at con- stant values, O 2 consumption integrated over the entire period from 0-50 Ma is 34.1×10 18 , -7.9×10 18 , or -38.8×10 18 moles O 2 for burial fluxes of 0.22×10 18 , 124 0.67×10 18 , and 1×10 18 moles S Ma −1 respectively (Figure 5.8a). Integrated over the period from 0-15 Ma, O 2 consumption is 19.3×10 18 , 6.7×10 18 , or -2.6×10 18 moles O 2 for burial fluxes of 0.22×10 18 , 0.67×10 18 , and 1×10 18 moles S Ma −1 respectively (Figure 5.8b). The results of the S cycle model imply changes in atmospheric O 2 . Although they notably do not include consideration of the organic carbon cycle, which may alsoinfluencepO 2 ,theresultsoftheScyclemodelimplyPaleocenepO 2 valuesof23 to 33% for the combinations of parameter values that produce net O 2 consumption when integrated over the entire period from 0 to 50 Ma (Figure 5.8a). Arguably, the combinations of parameter values that produce net O 2 consumption together with CO 2 production (i.e. a δ 34 S of riverine input > +0h and Δ pyrite−seawater ≥ -40h) are the most geologically reasonable given that the δ 34 S of riverine input is thought to be between +3 and +8h (Paytan et al., 1998; Kurtz et al., 2003), and the average Δ pyrite−seawater is thought to be -43± 2h (Wu et al., 2010). Considering only the last 15 Ma, which is when the (Li & Elderfield, 2013) model suggests that sulfide oxidation acted to balance enhanced CO 2 consumption as a result of silicate weathering, the parameter combinations that produce net sulfide oxidation imply Miocene pO 2 values of less than 30% and would produce sufficient CO 2 to balance plausible increases in silicate weathering fluxes (Figure 5.8b). This is consistent with the analysis as described in Figure 5.4. 125 0.50 0.49 0.51 0.52 100 150 200 0.51 0.49 0.53 0.55 1.20 0.80 1.60 2.00 2.40 1.20 0.80 1.60 2.00 2.40 Time (Ma) Relative Input Flux DIC sedimentary / DIC total Time (Ma) 100 150 200 Figure 5.2: Conceptual model of the timescale of CO 2 release from cou- pledsulfideoxidationandcarbonatedissolutionassociatedwiththecon- trastingtimescalesofsulfideandcarbonateburial. CO 2 release(redportion of the graph) occurs when the ratio of DIC derived from sedimentary carbonates (DIC sedimentary ) to DIC total is above 0.5, which is the value associated with the dis- solution of carbonates by carbonic acid and is pCO 2 neutral (i.e. the precipitation of carbonates from a fluid with this composition would have no effect pCO 2 ). A, The timescale of CO 2 release (solid green line) associated with an instantaneous doubling of the input fluxes (dashed blue line) from carbonate weathering by car- bonic acid and sulfuric acid. B, The timescale of CO 2 release (solid green line) associated with a doubling of the input fluxes (dashed blue line) from carbonate weathering by carbonic acid and sulfuric acid that occurs progressively over 40 Myr. 126 0 10 20 30 40 50 Age (Ma) 0.4 0.6 0.8 1.0 Silicate weathering 0.7075 0.7080 0.7085 0.7090 0.7095 87 Sr/ 86 Sr 0.0 0.5 1.0 Sediment weathering 0.4 0.6 0.8 1.0 187 Os/ 188 Os 0.5 1.0 1.5 0 10 20 30 40 50 Age (Ma) p CO 2 (ppm) 1000 2000 200 500 3000 Net drawdown of CO 2 from mountain uplift Figure 5.3: Results of the isotope mass balance model calculations of the relative changes in silicate weathering, sedimentary rock weather- ing, and the net drawdown of CO 2 by uplift compared with the 87 Sr/ 86 Sr (McArthur et al., 2001), 187 Os/ 188 Os (Klemm et al., 2005), and alkenone-based atmospheric pCO 2 records (error bars reflect range of pCO 2 values for different proxy calibrations; Pagani et al. 2005; Zhang et al. 2013). Values of silicate weath- ering, sedimentary rock weathering, and the net drawdown CO 2 by uplift are all relative to modern fluxes with the width of each of the curves reflecting the prop- agated uncertainty (1 s.d.) associated with the model input parameters. The shaded period highlights a major increase in the net weathering CO 2 sink that is associated with a drop in atmospheric pCO 2 (Pagani et al., 2005; Zhang et al., 2013). 127 0 7.5 45 30 15 0 O 2 consumed (10 18 moles) CO 2 released (10 18 moles) 50% Li and Elderfield 2013 estimate 15 35% 30% 25% 21% pO 2 at 15 Ma Watson et al. 1978 Fire limit Phanerozoic Max Berner et al. 2000 Modern Atmosphere Range supported by Wildman et al. 2004 25% 10% Figure5.4: Diagramillustratingthetrade-offbetweenCO 2 releaseandO 2 consumption as a result of sulfide oxidation (solid black line: ratio of CO 2 release to O 2 consumption for sulfide oxidation). The dashed, colored horizontal lines represent the CO 2 release required to fully compensate for a 10% (blue), 25% (green), or 50% (orange) increase in continental silicate weathering that occurred linearly over the last 15 Ma assuming a modern flux of 3.5×10 18 moles/Ma (see Methods). The 25% increase corresponds to the increase predicted by the Li & Elderfield (2013) mass balance model and implies pO 2 at 15 Ma of approximately 25%. Note that the estimate for the maximum possible atmosphere O 2 content of 25%, presented by Watson et al. (1978), is at odds with modeling (Berner, 2000) and experimental (Wildman et al., 2004) evidence for higher maximum O 2 contents. 128 Time (Ma) 0 10 20 30 40 50 S Flux (10 18 mol S / Ma) 0.0 1.0 2.0 3.0 long-term pyrite burial flux (Holser et al. 1988) Total riverine S flux from Li and Elderfield 10 model Pyrite derived riverine S flux from Li and Elderfield 10 model Calculated pyrite S burial flux for δ 34 S riverine = 0‰ and Δpyrite-seawater = -30 to -50‰ Calculated pyrite S burial flux for δ 34 S riverine = 5‰ and Δpyrite-seawater = -30 to -50‰ Calculated pyrite S burial flux for δ 34 S riverine = 10‰ and Δpyrite-seawater = -30 to -50‰ Figure 5.5: Riverine input fluxes and calculated pyrite burial fluxes from 0-50 Ma for constantδ 34 S values of riverine input and Δ pyrite−seawater . Solid and dashed black lines indicate the total and pyrite-derived input fluxes of sulfate from the Li & Elderfield (2013) model respectively. The green shape indicates the range of calculated pyrite burial fluxes for experiments where the δ 34 S value of riverine S flux was equal to 10h and Δ pyrite−seawater was varied between -30 and -50h. The red shape indicates the range of calculated pyrite burial fluxes for experiments where the δ 34 S value of riverine S flux was equal to 5h and Δ pyrite−seawater was varied between -30 and -50h. The blue shape indicates the range of calculated pyrite burial fluxes for experiments where the δ 34 S value of riverine was equal to 0h and Δ pyrite−seawater was varied between -30 and -50h. The yellow bar indicates the range of pyrite burial fluxes estimated from the pyrite content of marine sediments by Holser et al. (1988). 129 -250 -200 -150 -100 -50 0 0 5 10 15 20 25 30 Time (Ma) Δpyrite-seawater (‰) Average Δpyrite-seawater (+/- 1 σ) (Wu et al. 2010) Max Δpyrite-seawater (Wu et al. 2010) Calculated Δpyrite-seawater for pyrite burial flux = 0.22*10 18 mol S Ma -1 and δ 34 S riverine = 0 to +10‰ Calculated Δpyrite-seawater for pyrite burial flux = 0.67*10 18 mol S Ma -1 and δ 34 S riverine = 0 to +10‰ Calculated Δpyrite-seawater for pyrite burial flux = 1.00*10 18 mol S Ma -1 and δ 34 S riverine = 0 to +10‰ Figure 5.6: Calculated Δ pyrite−seawater values from 0-30 Ma for constantδ 34 S values of riverine input and pyrite burial fluxes. The solid and dashed black lines indicate the average and minimum Δ pyrite−seawater values for the Phanerozoic basedontheisotopicoffsetbetweencoevalmarinesulfateandsulfidemineralsfrom Wu et al. (2010). The thickness of the solid line reflects one standard deviation of the average Δ pyrite−seawater from Wu et al. (2010). The green shape indicates the range of calculated Δ pyrite−seawater values for experiments where the pyrite burial flux was equal to 1×10 18 mol S Ma −1 and the δ 34 S value of riverine was varied between 0 and 10h. The blue shape indicates the range of calculated Δ pyrite−seawater values for experiments where the pyrite burial flux was equal to 0.67×10 18 mol S Ma −1 and theδ 34 S value of riverine was varied between 0 and 10 h. The orange shape indicates the range of calculated Δ pyrite−seawater values for experiments where the pyrite burial flux was equal to 0.22×10 18 mol S Ma −1 and the δ 34 S value of riverine was varied between 0 and 10h. 130 0 5 10 15 20 25 30 -20 -10 0 10 20 Time (Ma) δ 34 S riverine input (‰) Estimated range modern δ 34 S riverine value (Paytan et al. 1998) Calculated δ 34 S riverine for pyrite burial flux = 0.22*10 18 mol S Ma -1 and Δpyrite-seawater = -30 to -50‰ Calculated δ 34 S riverine for pyrite burial flux = 0.67*10 18 mol S Ma -1 and Δpyrite-seawater = -30 to -50‰ Calculated δ 34 S riverine for pyrite burial flux = 1.00*10 18 mol S Ma -1 and Δpyrite-seawater = -30 to -50‰ Figure 5.7: Calculated δ 34 S values of riverine input from 0-30 Ma for constant Δ pyrite−seawater values and pyrite burial fluxes. The solid black line on the Y-axis indicates the range of estimates for the modernδ 34 S value of riverine input from Paytan et al. (1998); Kurtz et al. (2003). The green shape indicates the range of calculated δ 34 S values of riverine input for experiments where the pyrite burial flux was equal to 1×10 18 mol S Ma −1 and Δ pyrite−seawater was varied between -30 and -50h. The blue shape indicates the range of calculated δ 34 S values of riverine input for experiments where the pyrite burial flux was equal to 0.67×10 18 mol S Ma −1 and Δ pyrite−seawater was varied between -30 and -50h. The orange shape indicates the range of calculatedδ 34 S values of riverine input for experiments where the pyrite burial flux was equal to 0.22×10 18 mol S Ma −1 and Δ pyrite−seawater was varied between -30 and -50h. 131 10/-50 10/-30 10/-40 5/-50 5/-40 5/-30 0/-50 0/-40 0/-30 0.22 0.67 1 Integrated O 2 consumption (10 18 mols) Integrated CO 2 production (10 18 mols) 40 30 20 10 0 -10 -20 -30 -40 -20 -10 0 10 20 -30 30 25 20 15 10 5 0 -5 -10 -5 0 5 10 15 -10 Integrated CO 2 production (10 18 mols) 35% 30% 25% 21% 30% 25% 21% paleo-pO 2 δ 34 S riverine input /Δpyrite-seawater (permil) Sulfide burial (10 18 mols/Ma) b. 0-15 Ma a. 0-50 Ma 50% 25% 10% Figure 5.8: Integrated O 2 consumption and C) 2 production implied by the S cycle inverse model for different input parameters. The O 2 con- sumption values are presented in molar units. Importantly, the implied pO 2 values ignore other processes affecting the O 2 budget (e.g. changes in the organic carbon cycle). Negative values of both CO 2 and O 2 consumption reflect net sulfide burial. Independent estimates of the δ 34 S of riverine input (Paytan et al., 1998; Kurtz et al., 2003) and the average Δ pyrite−seawater (Wu et al., 2010) correspond with the parameter combinations that produce net sulfide oxidation (i.e. positive values) and are marked with stars. A) The model results integrated from 0-50 Ma. B) The model results integrated from 0-15 Ma following the results of the Li & Elder- field (2013) model, which suggests that sulfide oxidation acted at least in part to balance enhanced CO 2 consumption over this time period. The dashed, colored horizontal lines represent the CO 2 release required to fully compensate for a 10% (blue), 25% (green), or 50% (orange) increase in continental silicate weathering that occurred linearly over the last 15 Ma assuming a modern flux of 3.5×10 18 mol S Ma −1 (see section 5.3). The 25% increase corresponds to the increase predicted by the Li & Elderfield (2013) mass balance model. 132 Chapter 6 Conlusions The chemical transformation of rocks and minerals at Earth’s surface drives the formation of soils, regulates Earth’s climate over geologic timescales, and may offer a geo-engineering strategy to combat anthropogenic CO 2 emissions. Despite its general importance, there remain many outstanding questions related to chem- ical weathering processes and how they affect global biogeochemical cycles. The results of this dissertation provide novel insights into several outstanding ques- tions. Namely, 1) how do biologic processes affect environmental reaction rates? 2) does co-variation in hydrologic and geomorphic processes amplify or dampen variations in chemical weathering fluxes? and 3) how does the weathering of trace sulfide minerals affect acid budgets measured at local and global scales? The key findings of each chapter are briefly reviewed below. 6.1 Microbe-mineral interactions In chapter 2, laboratory dissolution experiments were used to determine the effects of siderophores (organic Fe-binding ligands) on the rates of olivine ((Mg,Fe) 2 SiO 4 ) dissolution. In natural environments, Fe is typically present at extremely low concentrations in solution, necessitating the production of siderophores by microorganisms to maintain an adequate nutrient supply. The experimental results confirm that the rates of dissolution of a wide variety of Fe- silicate minerals are accelerated by siderophores. In the case of olivine, I hypoth- esize that the observed acceleration in the dissolution rate is due to the removal of an inhibitory Fe-oxyhydroxide coating. Extrapolating the results to natural systems suggests that the extent to which dissolution rates are accelerated in nat- ural systems is controlled by complex feedbacks between microbial growth, ligand production, and mineral dissolution. 133 6.2 Hydrologic controls on chemical weathering In chapter 3, variations in the concentrations and concentration ratios of solutes in rivers draining the Peruvian Andes and Amazon were used to examine how hydrologic processes affect chemical weathering fluxes in different erosional environments. The hydrologic control on chemical weathering, expressed as the concentration-runoff relationship, was found to vary predictably between differ- ent erosional regimes. This results suggest that co-variation in hydrologic and erosional processes acts to amplify the differences in chemical weathering fluxes between rapidly and slowly eroding environments. Consequently, tectonic uplift may play a key role in the feedback between chemical weathering and climatic conditions. 6.3 Sulfide oxidation in the Andes/Amazon In chapter 4, a myriad of geochemical proxies were utilized to determine what proportion of chemical weathering is driven by sulfuric acid in the Andes/Amazon system and how this proportion changes between rapidly and slowly eroding environments. Chemical weathering by sulfuric acid can release CO 2 into the ocean/atmosphere system, but its importance and environmental controls are poorly understood. The results of this study show that sulfuric acid weather- ing dominates in the rapidly eroding Andes Mountains where it drives net CO 2 release. This contrasts with the slowly eroding Amazonian foreland floodplain, where there is net CO 2 consumption. 6.4 The Cenozoic isotope-weathering paradox The results of chapter 4 suggest that rapidly eroding environments release CO 2 into the ocean/atmosphere system as a result of sulfuric acid weathering. This suggests that major mountain uplift events in Earth’s history may have had a more complex effect of Earth’s carbon cycle than previously considered. In chapter 5, this hypothesis is explored using the Cenozoic period as a test case. The Cenozoic period is characterized by multiple major mountain uplift events, large changes in the chemistry of the ocean, and global cooling. However, isotopic 134 proxies of weathering during the Cenozoic are inconsistent with the need for mass balance in the long-term carbon cycle. The results of chapter 5 provide a novel solution, tectonically-enhanced sulfuric acid weathering, to the Cenozoic "isotope- weathering" paradox. 135 Reference List Andermann C, Longuevergne L, Bonnet S, Crave A, Davy P, Gloaguen R (2012) Impact of transient groundwater storage on the discharge of Himalayan rivers. Nature Geoscience 5:127–132. Anderson SP, Dietrich WE, Torres R, Montgomery DR, Loague K (1997) Concentration-discharge relationships in runoff from a steep, unchanneled catch- ment. Water Resources Research 33:211. Andreae MO, Berresheim H, Bingemer H, Jacob DJ, Lewis BL, Li SM, Talbot RW (1990a) The atmospheric sulfur cycle over the Amazon Basin: 2. Wet season. Journal of Geophysical Research 95:16813–16824. Andreae M, Talbot R, Berresheim H, Beecher K (1990b) Precipitation chemistry in central Amazonia. Journal of Geophysical Research 95:16987–16999. Asner GP, Anderson CB, Martin RE, Knapp DE, Tupayachi R, Sinca F, Malhi Y (2014) Landscape-scale changes in forest structure and functional traits along an Andes-to-Amazon elevation gradient. Biogeosciences 11:843–856. Balci N, Shanks WC, Mayer B, Mandernack KW (2007) Oxygen and sulfur isotope systematics of sulfate produced by bacterial and abiotic oxidation of pyrite. Geochimica et Cosmochimica Acta 71:3796–3811. Ballew N (2011) Quantifying discharge in the Kosñipata-Madre de Dios river system, Peru Undergraduate thesis, University of Southern California. Beaulieu E, Goddéris Y, Donnadieu Y, Labat D, Roelandt C (2012) High sensitiv- ity of the continental-weathering carbon dioxide sink to future climate change. Nature Climate Change 2:346–349. Bellenger JP, Wichard T, Kustka aB, Kraepiel aML (2008) Uptake of molybde- numandvanadiumbyÂăaÂănitrogen-fixingsoilbacteriumusingÂăsiderophores. Nature Geoscience 1:243–246. Berges Ja, Mulholland M (2008) Enzymes and N Cycling. 136 Berner EK, Berner RA (2012) Global environment: water, air, and geochemical cycles Princeton University Press. Berner RA (2000) Isotope Fractionation and Atmospheric Oxygen: Implications for Phanerozoic O2 Evolution. Science (New York, N.Y.) 287:1630–1633. Berner R (1982) Burial of Organic Carbon and Pyrite Sulfur in the Modern Ocean: Its Geochemical and Environmental Signifigance. American Journal of Science 282:451–473. Berner R, Lasaga A, Garrels R (1983) The Carbonate-Silicate Geochemical Cycle And Its Effect on Atmospheric Carbon Dioxide Over the Past 100 Million Years. American Journal of Science 283:641–683. Berner RA (1989) Biogeochemical cycles of carbon and sulfur and their effect on atmospheric oxygen over phanerozoic time. Palaeogeography, Palaeoclimatology, Palaeoecology 75:97–122. Berner RA, Caldeira K (1997) The need for mass balance and feedback in the geochemical carbon cycle. Geology 25:955. Berraho EL, Lesueur D, Diem HG, Sasson A (1997) Iron requirement and siderophore production in Rhizobium ciceri during growth on an iron-deficient medium. World Journal of Microbiology and Biotechnology 13:501–510. Beven KJ (2011) Rainfall-runoff modelling: the primer John Wiley & Sons. Bickle MJ, Tipper ED, Galy A, Chapman H, Harris N (2015) On discrimination between carbonate and silicate inputs to Himalayan rivers. American Journal of Science 315:120–166. Bickle M (1996) Metamorphic decarbonation, silicate weathering and the long- term carbon cycle. Terra Nova 8:270–276. Bickle M, Bunbury J, Chapman H, Harris N, Fairchild I, Ahmad T (2003) Fluxes of Sr into the headwaters of the Ganges. Geochimica et cosmochimica acta 67:2567–2584. Bird LJ, Bonnefoy V, Newman DK (2011) Bioenergetic challenges of microbial iron metabolisms. Trends in Microbiology 19:330–340. Blum JD, Erel Y (1997) Rb-Sr isotope systematics of a granitic soil chronose- quence: The importance of biotite weathering. Geochimica et Cosmochimica Acta 61:3193–3204. 137 Bluth GJS, Kump LR (1994) Lithologic and climatologic controls of river chem- istry. Geochimica et Cosmochimica Acta 58:2341–2359. Bonell M, Gilmour D (1978) The development of overland flow in a tropical rainforest catchment. BouchezJ,vonBlanckenburgF,SchuesslerJa(2013) Modelingnovelstableisotope ratios in the weathering zone. American Journal of Science 313:267–308. Bouchez J, Beyssac O, Galy V (2010) Oxidation of petrogenic organic carbon in the Amazon floodplain as a source of atmospheric CO2. Geology 38:255–258. Bouchez J, Gaillardet J, Lupker M, Louvat P, France-Lanord C, Maurice L, Armi- jos E, Moquet JS (2012) Floodplains of large rivers: Weathering reactors or simple silos? Chemical Geology 332-333:166–184. Boy J, Valarezo C, Wilcke W (2008) Water flow paths in soil control element exports in an Andean tropical montane forest. European Journal of Soil Sci- ence 59:1209–1227. Bradley DN, Tucker GE (2013) The storage time, age, and erosion hazard of laterally accreted sediment on the floodplain of a simulated meandering river. Journal of Geophysical Research: Earth Surface 118:1308–1319. Brantley SL, Lebedeva M, Bazilevskaya E (2014) Relating Weathering Fronts for Acid Neutralization and Oxidation to pCO2 and pO2 In Treatise on Geochem- istry: Second Edition, Vol. 6, pp. 327–352. Elsevier Ltd., 2 edition. Brantley SL, Holleran ME, Jin L, Bazilevskaya E (2013) Probing deep weathering intheShaleHillsCriticalZoneObservatory, Pennsylvania(USA):thehypothesis of nested chemical reaction fronts in the subsurface. Earth Surface Processes and Landforms 38:1280–1298. BrayAW,OelkersEH,BonnevilleS,Wolff-BoenischD,PottsNJ,FonesG,Benning LG (2015) The effect of pH, grain size and organic ligands on biotite weathering rates. Geochimica et Cosmochimica Acta 164:127–145. Budzikiewicz H (2010) Microbial Siderophores In Fortschritte Der Chemie Organ- ischer Naturstoffe / Progress in the Chemistry of Organic Natural Products, Vol. 92. Butman D, Raymond Pa (2011) Significant efflux of carbon dioxide from streams and rivers in the United States. Nature Geoscience 4:839–842. 138 Calmels D, Gaillardet J, Brenot A, France-Lanord C (2007) Sustained sulfide oxidation by physical erosion processes in the Mackenzie River basin: Climatic perspectives. Geology 35:1003–1006. Calmels D, Galy A, Hovius N, Bickle M, West AJ, Chen Mc, Chapman H (2011) Contribution of deep groundwater to the weathering budget in a rapidly eroding mountain belt, Taiwan. Earth and Planetary Science Letters 303:48–58. Calvaruso C, Turpault MP, Frey-Klett P (2006) Root-Associated Bacteria Con- tribute to Mineral Weathering and to Mineral Nutrition in Trees: a Budgeting Analysis. Applied and Environmental Microbiology 72:1258–1266. Carlotto Caillaux VS, Rodriguez G, Fernando W, Roque C, Dionicio J, Chávez R (1996) Geología de los cuadrángulos de Urubamba y Calca Technical report, Instituto Geológica Nacional, Lima, Peru. Carrasco N, Kretzschmar R, Pesch ML, Kraemer SM (2007) Low concentrations of surfactants enhance siderophore-promoted dissolution of goethite. Environ- mental science & technology 41:3633–8. Chadwick O, Derry L, Vitousek P (1999) Changing sources of nutrients during four million years of ecosystem development. Nature 397:491–497. Cheah SF, Kraemer SM, Cervini-Silva J, Sposito G (2003) Steady-state dissolution kinetics of goethite in the presence of desferrioxamine B and oxalate ligands: Implications for the microbial acquisition of iron. Chemical Geology 198:63–75. Chen Y, Brantley SL (2000) Dissolution of forsteritic olivine at 65C and 2 < pH < 5. Chemical Geology 165:267–281. Christophersen N, Neal C, Hooper RP, Vogt R, Andersen S (1990) Mod- elling streamwater chemistry as a mixture of soilwater end-members âĂŤ A Step Towards Second-Generation Acidification Models. Journal of Hydrol- ogy 116:307–320. Clark KE, Hilton RG, West AJ, Malhi Y, Gröcke DR, Bryant CL, Ascough PL, Robles Caceres A, New M (2013) New views on âĂIJoldâĂİ carbon in the Ama- zon River: Insight from the source of organic carbon eroded from the Peruvian Andes. Geochemistry, Geophysics, Geosystems 14:1644–1659. ClarkKE,TorresMA,WestAJ,HiltonRG,NewM,HorwathAB,FisherJB,Rapp JM, Robles Caceres A, Malhi Y (2014) The hydrological regime of a forested tropicalAndeancatchment. Hydrology and Earth System Sciences18:5377–5397. 139 Clow DW, Mast MA (2010) Mechanisms for chemostatic behavior in catch- ments: Implications for CO2 consumption by mineral weathering. Chemical Geology 269:40–51. Das A, Chung CH, You CF (2012) Disproportionately high rates of sulfide oxida- tion from mountainous river basins of Taiwan orogeny: Sulfur isotope evidence. Geophysical Research Letters 39:1–6. Daval D, Testemale D, Recham N, Tarascon JM, Siebert J, Martinez I, Guyot F (2010) Fayalite (Fe2SiO4) dissolution kinetics determined by X-ray absorption spectroscopy. Chemical Geology 275:161–175. Dellinger M, Gaillardet J, Bouchez J, Calmels D, Galy V, Hilton RG, Louvat P, France-Lanord C (2014) Lithium isotopes in large rivers reveal the cannibalistic nature of modern continental weathering and erosion. Earth and Planetary Science Letters 401:359–372. Dellinger M, Gaillardet J, Bouchez J, Calmels D, Louvat P, Dosseto A, Gorge C, Alanoca L, Maurice L (2015) Riverine Li isotope fractionation in the Amazon River basin controlled by the weathering regimes. Geochimica et Cosmochimica Acta . Derry LA, Kurtz AC, Ziegler K, Chadwick OA (2005) Biological control of terres- trial silica cycling and export fluxes to watersheds. Nature 433:728–731. Dessert C, Dupre B, Gaillardet J, Francois LM, Allegre CJ (2003) Basalt weather- ing laws and the impact of basalt weathering on the global carbon cycle. Chem- ical Geology 202:257–273. Detmers J, Brüchert V, Habicht KS, Kuever J (2001) Diversity of sulfur iso- tope fractionations by sulfate reducing prokaryotes. Appl. Environm. Micro- biol. 67:888–894. Doctor D, Kendall C (2008) Carbon isotope fractionation of dissolved inorganic carbon (DIC) due to outgassing of carbon dioxide from a headwater stream. Hydrological Processes 2423:2410–2423. Dokulil MT (1994) Environmental control of phytoplankton productivity in tur- bulent turbid systems. Hydrobiologia 289:65–72. Dosseto A, Bourdon B, Gaillardet J, Maurice-Bourgoin L, Allègre CJ (2006) Weathering and transport of sediments in the Bolivian Andes: Time constraints from uranium-series isotopes. Earth and Planetary Science Letters 248:759–771. Dunne T (1978) Field studies of hillslope flow processes. Hillslope hydrol- ogy 227:293. 140 Edmond J (1992) Himalayan Tectonics, Weathering Processes, and the Strontium Isotope Record in Marine Limestones. Science (New York, N.Y.) 258:1594–1597. Eiriksdottir ES, Gislason SR, Oelkers EH (2013) Does temperature or runoff con- trol the feedback between chemical denudation and climate? Insights from NE Iceland. Geochimica et Cosmochimica Acta 107:65–81. Elsenbeer H, Lack A, Cassel K (1995) Chemical fingerprints of hydrological com- partments and flow paths and La Cuenca, western Amazonia. Water Resources Research 31:3051–3058. Essén Sa, Bylund D, Holmström SJM, Moberg M, Lundström US (2006) Quan- tification of hydroxamate siderophores in soil solutions of podzolic soil profiles in Sweden. Biometals 19:269–82. Evans MJ, Derry La, Anderson SP, France-Lanord C (2002) Hydrothermal source of radiogenic Sr to Himalayan rivers. Geology 29:803–806. Evers A, Hancock R, Martell A, Motekaitis R (1989) Metal Ion Recognition in Lig- ands with Negatively Charged Oxygen Donor Groups. Complexation of Fe(III), Ga(III),In(III),Al(III),andOtherHighlyChargedMetalIons. Inorganic Chem- istry 28:2189–2195. Falkowski PG, Katz ME, Milligan AJ, Fennel K, Cramer BS, Aubry MP, Berner Ra, Novacek MJ, Zapol WM (2005) The rise of oxygen over the past 205 mil- lion years and the evolution of large placental mammals. Science (New York, N.Y.) 309:2202–2204. Farkas E, Enyedy Ea, Zékány L, Deák G (2001) Interaction between iron(II) and hydroxamic acids: oxidation of iron(II) to iron(III) by desferrioxamine B under anaerobic conditions. Journal of inorganic biochemistry 83:107–14. Fekete F, Spence J, Emery T (1983) Siderophores Produced by Nitrogen-Fixing Azotobacter vinelandii OP in Iron-Limited Continuous Culture. Applied and Environmental Microbiology 46:1297–1300. Ferrier KL, Kirchner JW (2008) Effects of physical erosion on chemical denuda- tion rates: A numerical modeling study of soil-mantled hillslopes. Earth and Planetary Science Letters 272:591–599. Fischer C, Arvidson RS, Lüttge A (2012) How predictable are dissolution rates of crystalline material? Geochimica et Cosmochimica Acta 98:177–185. FranceâĂŘLanord C, Derry L (1997) Organic carbon burial forcing of the carbon cycle from Himalayan erosion. Nature 390:65–67. 141 Furch K (1984) Water chemistry of the Amazon basin: the distribution of chemical elements among freshwaters In The Amazon, pp. 167–199. Springer. Gabet EJ, Mudd SM (2009) A theoretical model coupling chemical weathering rates with denudation rates. Geology 37:151–154. Gabet EJ, Edelman R, Langner H (2006) Hydrological controls on chemical weath- ering rates at the soil-bedrock interface. Geology 34:1065. Gaillardet J, Dupre B, Allegre C, Négrel P (1997) Chemical and physical denuda- tion in the Amazon River Basin. Chemical Geology 142:141–173. Gaillardet J, Dupré B, Louvat P, Allegre CJ (1999) Global silicate weathering and CO2 consumption rates deduced from the chemistry of large rivers. Chemical Geology 159:3–30. Gaillardet J, Rad S, Rivé K, Louvat P, Gorge C, Allègre CJ, Lajeunesse E (2011) Orography-driven chemical denudation in the lesser antilles: Evidence for a new feed-back mechanism stabilizing atmospheric CO 2. American Journal of Science 311:851–894. GalyA,France-LanordC(1999) WeatheringprocessesintheGanges-Brahmaputra basin and the riverine alkalinity budget. Chemical Geology 159:31–60. Galy V, Beyssac O, France-Lanord C, Eglinton T (2008) Recycling of graphite dur- ing Himalayan erosion: a geological stabilization of carbon in the crust. Science (New York, N.Y.) 322:943–5. GalyV,France-LanordC,BeyssacO,FaureP,KudrassH,PalholF(2007) Efficient organic carbon burial in the Bengal fan sustained by the Himalayan erosional system. Nature 450:407–10. Ganor J, Roueff E, Erel Y, Blum JD (2005) The dissolution kinetics of a granite and its minerals - Implications for comparison between laboratory and field dissolution rates. Geochimica et Cosmochimica Acta 69:607–621. Gautier JM, Oelkers EH, Schott J (2001) Are quartz dissolution rates proportional to B.E.T. surface areas? Geochimica et Cosmochimica Acta 65:1059–1070. GeorgRB,ReynoldsBC,WestAJ,BurtonKW,HallidayAN(2007) Siliconisotope variations accompanying basalt weathering in Iceland. Earth and Planetary Science Letters 261:476–490. Georg R, a.J. West, Vance D, Newman K, a.N. Halliday (2013) Is the marine osmium isotope record a probe for CO2 release from sedimentary rocks? Earth and Planetary Science Letters 367:28–38. 142 Gibbs R (1967) The Geochemistry of the Amazon River System: Part I. The Factors that Control the Salinity and the Composition and Concentration of the Suspended Solids. Geological Society of America Bulletin 78:1203–1232. Girardin CaJ, Espejob JES, Doughty CE, Huasco WH, Metcalfe DB, Durand- Baca L, Marthews TR, Aragao LEOC, Farfan-Rios W, García-Cabrera K, Halla- day K, Fisher JB, Galiano-Cabrera DF, Huaraca-Quispe LP, Alzamora-Taype I, Eguiluz-Mora L, Salinas-Revilla N, Silman MR, Meir P, Malhi Y (2013) Produc- tivity and carbon allocation in a tropical montane cloud forest in the Peruvian Andes. Plant Ecology & Diversity pp. 1–17. Godsey S, Kirchner J, Clow D (2009) ConcentrationâĂŞdischarge relation- ships reflect chemostatic characteristics of US catchments. Hydrological Pro- cesses 1864:1844–1864. Graham RC, Rossi AM, Hubbert KR (2010) Rock to regolith conversion: Produc- ing hospitable substrates for terrestrial ecosystems. GSA Today 20:4–9. Gröger J, Franke J, Hamer K, Schulz HD (2009) Quantitative recovery of elemen- tal sulfur and improved selectivity in a chromium-reducible sulfur distillation. Geostandards and Geoanalytical Research 33:17–27. Guyot JL, Jouanneau JM, Soares L, Boaventura GR, Maillet N, Lagane C (2007) Clay mineral composition of river sediments in the Amazon Basin. Catena 71:340–356. Haack EA, Johnston CT, Maurice Pa (2008) Mechanisms of siderophore sorption to smectite and siderophore-enhanced release of structural Fe3+. Geochimica et Cosmochimica Acta 72:3381–3397. Habicht KS, Gade M, Thamdrup B, Berg P, Canfield DE (2002) Calibration of sulfate levels in the archean ocean. Science (New York, N.Y.) 298:2372–4. Hahm WJ, Riebe CS, Lukens CE, Araki S (2014) Bedrock composition regu- lates mountain ecosystems and landscape evolution. Proceedings of the National Academy of Sciences of the United States of America 111:3338–43. Hanlon DP, Watt DS, Westhead EW (1966) The interaction of divalent metal ions with tris buffer in dilute solution. Analytical biochemistry 16:225–233. Hartmann J, Jansen N, Dürr HH, Kempe S, Köhler P (2009) Global CO2- consumption by chemical weathering: What is the contribution of highly active weathering regions? Global and Planetary Change 69:185–194. 143 Hem J (1972) Chemical Factors that Influence the Availability of Iron and Manganese in Aqueous Systems. Bulletin of the Geological Society of Amer- ica 83:443–450. Hercod DJ, Brady PV, Gregory RT (1998) Catchment-scale coupling between pyrite oxidation and calcite weathering. Chemical geology 151:259–276. HilleyG,ChamberlainC,MoonS,PorderS,WillettS(2010) Competitionbetween erosion and reaction kinetics in controlling silicate-weathering rates. Earth and Planetary Science Letters 293:191–199. Hilton RG, Gaillardet J, Calmels D, Birck JL (2014) Geological respiration of a mountain belt revealed by the trace element rhenium. Earth and Planetary Science Letters 403:27–36. Hodson ME (2003) The influence of Fe-rich coatings on the dissolution of anorthite at pH 2.6. Geochimica et Cosmochimica Acta 67:3355–3363. Holmström S, Lundström U (2004) Siderophores in forest soil solution. Biogeo- chemistry 71:247–258. Holser WT,SchidlowskiM, Mackenzie FT, Maynard JB(1988) Geochemical cycles of carbon and sulfur. Chemical Cycles in the Evolution of the Earth pp. 105–173. Homann VV, Edwards KJ, Webb EA, Butler A (2009) Siderophores of Marinobac- ter aquaeolei: petrobactin and its sulfonated derivatives. Biometals 22:565–571. Horita J, Zimmermann H, Holland H (2002) Chemical evolution of seawater during the Phanerozoic: Implications from the record of marine evaporites. Geochimica et Cosmochimica Acta 66:3733–3756. Huasco WH, a.J. Girardin C, Doughty CE, Metcalfe DB, Baca LD, Silva-Espejo JE, Cabrera DG, Aragão LE, Davila AR, Marthews TR, Huaraca-Quispe LP, Alzamora-Taype I, Mora LE, Farfán-Rios W, Cabrera KG, Halladay K, Salinas- Revilla N, Silman MR, Meir P, Malhi Y (2014) Seasonal production, allocation and cycling of carbon in two mid-elevation tropical montane forest plots in the Peruvian Andes. Plant Ecology & Diversity 7:125–142. Huh Y, Birck JL, Allègre CJ (2004a) Osmium isotope geochemistry in the Macken- zie River basin. Earth and Planetary Science Letters 222:115–129. HuhY,ChanLH,ChadwickOa(2004b) Behavioroflithiumanditsisotopesduring weathering of Hawaiian basalt. Geochemistry, Geophysics, Geosystems 5:1–22. INGEMMET (2013) GEOCATMIN - Geologia integrada por proyectos regionales Technical report, Instituto Geológico Minero Metalúrgico. 144 Jacobson AD, Blum JD, Walter LM (2002) Reconciling the elemental and Sr isotopecompositionofHimalayanweatheringfluxes: insightsfromthecarbonate geochemistryofstreamwaters. Geochimica et Cosmochimica Acta66:3417–3429. Jacobson A, Blum J, Chamberlain C, Craw D, Koons P (2003) Climatic and tectonic controls on chemical weathering in the New Zealand Southern Alps. Geochimica et Cosmochimica Acta 67:29–46. Jarvis A, Reuter HI, Nelson A, Guevara E, Others (2008) Hole-filled SRTM for the globe Version 4. available from the CGIAR-CSI SRTM 90m Database (http://srtm. csi. cgiar. org) . Johnson N, Likens G, Bormann F, Fisher D, Pierce R (1969) A Working Model for the Variation in Stream Water Chemistry at the Hubbard Brook Experimental Forest, New Hampshire. Water Resources Research 5:1353–1363. Katz M, Wright J, Miller K, Cramer B, Fennel K, Falkowski P (2005) Biological overprint of the geological carbon cycle. Marine Geology 217:323–338. Keech AR, West aJ, Pett-Ridge JC, Henderson GM (2013) Evaluating U-series tools for weathering rate and duration on a soil sequence of known ages. Earth and Planetary Science Letters 374:24–35. Kerrick DM, Caldeira K (1998) Metamorphic CO2 degassing from orogenic belts. Chemical Geology 145:213–232. Kim H, Bishop JK, Dietrich WE, Fung IY (2014) Process dominance shift in solute chemistry as revealed by long-term high-frequency water chemistry observations of groundwater flowing through weathered argillite underlying a steep forested hillslope. Geochimica et Cosmochimica Acta 140:1–19. Kirchner JW, Neal C (2013) Universal fractal scaling in stream chemistry and its implications for solute transport and water quality trend detection. Proceedings of the National Academy of Sciences . Kirchner JW (2015) Aggregation in environmental systems: seasonal tracer cycles quantify young water fractions, but not mean transit times, in spa- tially heterogeneous catchments. Hydrology and Earth System Sciences Dis- cussions 12:3059–3103. Klemm V, Levasseur S, Frank M, Hein JR, Halliday AN (2005) Osmium isotope stratigraphy of a marine ferromanganese crust. Earth and Planetary Science Letters 238:42–48. 145 Köhler P, Hartmann J, Wolf-Gladrow DA (2010) Geoengineering potential of artificially enhanced silicate weathering of olivine. Proceedings of the National Academy of Sciences 107:20228. Kurtz AC, Kump LR, Arthur MA, Zachos JC, Paytan A (2003) Early Cenozoic decoupling of the global carbon and sulfur cycles. Paleoceanography 18:1–14. Kurtz AC, Lugolobi F, Salvucci G (2011) Germanium-silicon as a flow path tracer: Application to the Rio Icacos watershed. Water Resources Research 47:W06516. Lambs L, Horwath A, Otto T, Julien F, Antoine PO (2012) Isotopic values of the Amazon headwaters in Peru: comparison of the wet upper Río Madre de Dios watershed with the dry Urubamba-Apurimac river system. Rapid communica- tions in mass spectrometry 26:775–84. LaRowe DE, Van Cappellen P (2011) Degradation of natural organic matter: A thermodynamic analysis. Geochimica et Cosmochimica Acta 75:2030–2042. Larsen IJ, Montgomery DR, Greenberg HM (2014) The contribution of mountains to global denudation. Geology pp. 1–4. Lawrence C, Harden J, Maher K (2014) Modeling the influence of organic acids on soil weathering. Geochimica et Cosmochimica Acta 139:487–507. Lebedeva M, Fletcher R, Balashov V, Brantley S (2007) A reactive diffu- sion model describing transformation of bedrock to saprolite. Chemical Geol- ogy 244:624–645. Leleyter L, Probst J (1999) A new sequential extraction procedure for the speci- ation of particulate trace elements in river sediments. International Journal of Environmental Analytical Chemistry 73:109–128. Lemarchand E, Chabaux F, Vigier N, Millot R (2010) Lithium isotope system- atics in a forested granitic catchment (Strengbach, Vosges Mountains, France). Geochimica et Cosmochimica Acta 74:4612–4628. Lerman A, Wu L, Mackenzie FT (2007) CO2 and H2SO4 consumption in weath- ering and material transport to the ocean, and their role in the global carbon balance. Marine Chemistry 106:326–350. Levasseur S, Birck J, Allegre C (1999) The osmium riverine flux and the oceanic mass balance of osmium. Earth and Planetary Science Letters 174:7–23. Li DD, Jacobson AD, McInerney DJ (2014) A reactive-transport model for exam- ining tectonic and climatic controls on chemical weathering and atmospheric CO2 consumption in granitic regolith. Chemical Geology 365:30–42. 146 Li G, Elderfield H (2013) Evolution of carbon cycle over the past 100 million years. Geochimica et Cosmochimica Acta 103:11–25. Li G, Ji J, Chen J, Kemp DB (2009) Evolution of the Cenozoic carbon cycle: The roles of tectonics and CO2 fertilization. Global Biogeochemical Cycles 23:1–11. Li SL, Calmels D, Han G, Gaillardet J, Liu CQ (2008) Sulfuric acid as an agent of carbonate weathering constrained by δ13CDIC: Examples from Southwest China. Earth and Planetary Science Letters 270:189–199. Liermann L, Kalinowski B, Brantley S, Ferry J (2000) Role of bacterial siderophores in dissolution of hornblende. Geochimica et Cosmochimica Acta 64:587–602. Longinelli A, Edmond J (1983) Isotope geochemistry of the Amazon basin: a reconnaissance. Journal of Geophysical Research 88:3703–3717. Lu FH, Meyers WJ, a.a. Schoonen M (1997) Minor and trace element analyses on gypsum: an experimental study. Chemical Geology 142:1–10. Maher K (2010) The dependence of chemical weathering rates on fluid residence time. Earth and Planetary Science Letters 294:101–110. Maher K (2011) The role of fluid residence time and topographic scales in deter- mining chemical fluxes from landscapes. Earth and Planetary Science Let- ters 312:48–58. Maher K, Chamberlain CP (2014) Hydrologic Regulation of Chemical Weathering and the Geologic Carbon Cycle. Science (New York, N.Y.) 1502. Mayorga E, Aufdenkampe AK, Masiello CA, Krusche AV, Hedges JI, Quay PD, Richey JE, Brown TA (2005) Young organic matter as a source of carbon dioxide outgassing from Amazonian rivers. Nature 436:538–541. McArthur JM, Howarth RJ, Bailey TR (2001) Strontium Isotope Stratigraphy: LOWESS Version 3: Best Fit to the Marine SrâĂŘIsotope Curve for 0âĂŞ509 Ma and Accompanying LookâĂŘup Table for Deriving Numerical Age. The Journal of Geology 109:155–170. McGuire KJ, McDonnell JJ, Weiler M, Kendall C, McGlynn BL, Welker JM, Seib- ert J (2005) The role of topography on catchment-scale water residence time. Water Resources Research 41:1–14. McGuire KJ, McDonnell JJ (2006) A review and evaluation of catchment transit time modeling. Journal of Hydrology 330:543–563. 147 Meade RH (1985) Suspended sediment in the Amazon River and its tributaries in Brazil during 1982-84 Technical report, US Geological Survey,. Meister P (2013) Two opposing effects of sulfate reduction on carbonate pre- cipitation in normal marine, hypersaline, and alkaline environments. Geol- ogy 41:499–502. Mendez J, Guieu C, Adkins J (2010) Atmospheric input of manganese and iron to the ocean: Seawater dissolution experiments with Saharan and North American dusts. Marine Chemistry 120:34–43. Mendívil Echevarría S, Dávila Manrique D (1994) Geología de los cuadrángulos de Cuzco y Livitaca Technical report, Instituto Geológica Nacional, Lima, Peru. Meybeck M, Ragu A (1996) GEMS/water contribution to the global register of riverinputs.GEMS/WaterProgramme(UNEP/WHO/UNESCO). World Health Organization, Geneva, Switzerland . Millot R, Gaillardet J, Dupre B, Allegre C (2003) Northern latitude chemical weathering rates: clues from the Mackenzie River Basin, Canada. Geochimica et Cosmochimica Acta 67:1305–1329. Misra S, Froelich PN (2012) Lithium isotope history of Cenozoic seawater: changes in silicate weathering and reverse weathering. Science 335:818–23. Moon S, Chamberlain C, Hilley G (2014) New estimates of silicate weathering rates and their uncertainties in global rivers. Geochimica et Cosmochimica Acta 134:257–274. Moquet JS, Crave A, Viers J, Seyler P, Armijos E, Bourrel L, Chavarri E, Lagane C, Laraque A, Casimiro WSL, Pombosa R, Noriega L, Vera A, Guyot JL (2011) Chemical weathering and atmospheric/soil CO2 uptake in the Andean and Fore- land Amazon basins. Chemical Geology 287:1–26. Müller RD, Sdrolias M, Gaina C, Steinberger B, Heine C (2008) Long-term sea-level fluctuations driven by ocean basin dynamics. Science (New York, N.Y.) 319:1357–62. Negrel P, Allegre CJ, Dupre B, Lewin E (1993) Erosion sources determined by inversion of major and trace element ratios and strontium isotopic ratios in river water: The Congo Basin case. Earth and Planetary Science Letters 120:59–76. Neilands JB (1995) Siderophores: structure and function of microbial iron trans- port compounds. Journal of Biological Chemistry 270:26723. 148 Oelkers EH, Schott J (1998) Does organic acid adsorption affect alkali-feldspar dissolution rates? Chemical Geology 151:235–245. Oelkers E (2001) An experimental study of forsterite dissolution rates as a func- tion of temperature and aqueous Mg and Si concentrations. Chemical Geol- ogy 175:485–494. Olsen AA, Rimstidt JD (2008) Oxalate-promoted forsterite dissolution at low pH. Geochimica et Cosmochimica Acta 72:1758–1766. Pagani M, Zachos JC, Freeman KH, Tipple B, Bohaty S (2005) Marked decline in atmospheric carbon dioxide concentrations during the Paleogene. Science (New York, N.Y.) 309:600–3. Page WJ, Huyer M (1984) Derepression of the Azotobacter vinelandii siderophore system, using iron-containing minerals to limit iron repletion. Journal of bacte- riology 158:496–502. Paris G, Adkins JF, Sessions AL, Webb SM, Fischer WW (2014) Neoarchean carbonate-associated sulfate records positive âĹĘ33S anamalies. Sci- ence 346:739–742. Paris G, Sessions AL, Subhas AV, Adkins JF (2013) MC-ICP-MS measurement of δ34S and âĹĘ33S in small amounts of dissolved sulfate. Chemical Geol- ogy 345:50–61. Paytan A, Kastner M, Campbell D, Thiemens MH (1998) Sulfur isotopic compo- sition of cenozoic seawater sulfate. Science (New York, N.Y.) 282:1459–62. Playà E, Rosell L (2005) The celestite problem in gypsum Sr geochemistry: An evaluation of purifying methods of gypsiferous samples. Chemical Geol- ogy 221:102–116. Pogge von Strandmann PAE, Henderson GM (2014) The Li isotope response to mountain uplift. Geology 43:1–4. Pokrovsky O, Schott J (2000) Kinetics and mechanism of forsterite dissolution at 25 C and pH from 1 to 12. Geochimica et Cosmochimica Acta 64:3313–3325. Polsenaere P, Gwenael A (2012) Modelling CO2 degassing from small acidic rivers using water pCO2, DIC and d13C-DIC data. Geochimica et Cosmochimica Acta 91:220–239. Ponton C, West AJ, Feakins SJ, Galy V (2014) Leaf wax biomarkers in transit record river catchment composition. Geophysical Research Letters 41:6420–6427. 149 Porder S, Vitousek PM, Chadwick Oa, Chamberlain CP, Hilley GE (2007) Uplift, Erosion, and Phosphorus Limitation in Terrestrial Ecosystems. Ecosys- tems 10:159–171. Portenga EW, Bierman PR (2011) Understanding EarthâĂŹs eroding surface with 10Be. GSA Today 21:4–10. Powell PE, Cline GR, Reid CPP, Szaniszlo PJ (1980) Occurrence of hydroxamate siderophore iron chelators in soils. Nature . Räsänen M, Neller R, Salo J, Jungner H (1992) Recent and ancient fluvial deposition systems in the Amazonian foreland basin, Peru. Geological Maga- zine 129:293–306. Raymo M, Ruddiman W (1992) Tectonic forcing of late Cenozoic climate. Nature 359:117–122. Raymo M, Ruddiman W, Froelich P (1988) Influence of late Cenozoic mountain building on ocean geochemical cycles. Geology 16:649–653. Reeves D, Rothman DH (2013) Age dependence of mineral dissolution and pre- cipitation rates. Global Biogeochemical Cycles 27:n/a–n/a. ReidEA,ReidJS,MeierMM,DunlapMR,CliffSS,BroumasA,PerryK,MaringH (2003) CharacterizationofAfricandusttransportedtoPuertoRicobyindividual particle and size segregated bulk analysis. Rigsby Ca, Hemric EM, Baker Pa (2009) Late Quaternary Paleohydrology of the Madre de Dios River, southwestern Amazon Basin, Peru. Geomorphol- ogy 113:158–172. Rimstidt JD, Brantley SL, Olsen AA (2012) Systematic review of forsterite disso- lution rate data. Geochimica et Cosmochimica Acta 99:159–178. Römheld V, Marschner H (1986) Evidence for a specific uptake system for iron phytosiderophores in roots of grasses. Plant physiology 80:175–180. Sackett WM, Moore WS (1966) Isotopic variations of dissolved inorganic carbon. Chemical Geology 1:323–328. Salve R, Rempe DM, Dietrich WE (2012) Rain, rock moisture dynamics, and the rapid response of perched groundwater in weathered, fractured argillite under- lying a steep hillslope. Water Resources Research 48:1–25. Santelli CM, Welch SA, Westrich HR, Banfield JF (2001) The effect of Fe-oxidizing bacteria on Fe-silicate mineral dissolution. Chemical Geology 180:99–115. 150 Schulte P, van Geldern R, Freitag H, Karim A, Négrel P, Petelet-Giraud E, Probst A, Probst JL, Telmer K, Veizer J, a.C. Barth J (2011) Applications of stable water and carbon isotopes in watershed research: Weathering, carbon cycling, and water balances. Earth-Science Reviews 109:20–31. Shirvani M, Nourbakhsh F (2010) Desferrioxamine-B adsorption to and iron dis- solution from palygorskite and sepiolite. Applied Clay Science 48:393–397. Soetaert K, Hofmann AF, Middelburg JJ, Meysman FJR, Greenwood J (2007) The effect of biogeochemical processes on pH. Marine Chemistry 106:380–401. Stallard R, Edmond J (1981) Geochemistry of the Amazon 1. Precipitation chem- istry and the marine contribution to the dissolved load at the time of peak discharge. Journal of geophysical Research 86:9844–9858. Stallard R, Murphy S (2014) A Unified Assessment of Hydrologic and Biogeochem- icalResponsesinResearchWatershedsinEasternPuertoRicoUsingRunoffâĂŞ- Concentration Relations. Aquatic Geochemistry 20:115–139. Stallard RF (1974) Major Element Geochemistry of the Amazon River Ph.D. diss., Massachusetts Institute or Technology. Stocker T, Qin D, Plattner GK, Tignor M, Allen SK, Boschung J, Nauels A, Xia Y, Bex V, Midgley PM (2014) Climate change 2013: The physical science basis Cambridge University Press Cambridge, UK, and New York. Stumm W, Lee FG (1961) Oxygenation of ferrous iron. Industrial and engineering chemistry 53:143–146. Sugimori H, Kanzaki Y, Murakami T (2012) Relationships between Fe redistribu- tion and Po2 during mineral dissolution under low O2 conditions. Geochimica et Cosmochimica Acta 84:29–46. Sweet SP, Douglas LJ (1991) Effect of iron concentration on siderophore syn- thesis and pigment production by Candida albicans. FEMS microbiology let- ters 64:87–91. Tardy Y, Bustillo V, Roquin C, Mortatti J, Victoria R (2005) The Amazon. Bio- geochemistry applied to river basin management:: Part I. Hydro-climatology, hydrograph separation, mass transfer balances, stable isotopes, and modelling. Applied geochemistry 20:1746–1829. Tipper ET, Bickle MJ, Galy A, West AJ, Pomiès C, Chapman HJ (2006) The short term climatic sensitivity of carbonate and silicate weathering fluxes: Insight from seasonal variations in river chemistry. Geochimica et Cosmochimica Acta 70:2737–2754. 151 Tobón C, Sevink J, Verstraten JM (2004) Solute fluxes in throughfall and stemflow in four forest ecosystems in northwest Amazonia. Biogeochemistry 70:1–25. Torres MA, West AJ, Clark KE (2015) Geomorphic regime modulates hydrologic control of chemical weathering in the Andes-Amazon. In Review . Torres MA, West AJ, Li G (2014) Sulphide oxidation and carbonate dissolution as a source of CO2 over geological timescales. Nature 507:346–349. Turchyn AV, Tipper ET, Galy A, Lo JK, Bickle MJ (2013) Isotope evidence for secondary sulfide precipitation along the Marsyandi River, Nepal, Himalayas. Earth and Planetary Science Letters 374:36–46. Vargas Vilchez L, Hipolito Romero A (1998) Geología de los cuadrángulos de Río Pinquén, Pilcopata y Chontachaca. Hojas: 25-t, 26-t y 27-t Technical report, Instituto Geológica Nacional, Lima, Peru. Velbel M (1993) Formation of protective surface layers during silicate-mineral weathering under well-leached, oxidizing conditions. American Mineralo- gist 78:405–414. Vitousek P, Chadwick O, Matson P, Allison S, Derry L, Kettley L, Luers A, Meck- ingE,MonastraV,PorderS(2003) ErosionandtheRejuvenationofWeathering- derived Nutrient Supply in an Old Tropical Landscape. Ecosystems 6:762–772. Vraspir JM, Butler A (2009) Chemistry of Marine Ligands and Siderophores. Annual Review of Marine Science 1:43–63. Walker J, Hays P, Kasting J (1981) A negative feedback mechanism for the long- term stabilization of the Earth’s surface temperature. Journal of Geophysical Research 86:9776–9782. WatsonA,LovelockJE,MargulisL(1978) Methanogenesis,firesandtheregulation of atmospheric oxygen. Biosystems 10:293–298. West AJ (2012) Thickness of the chemical weathering zone and implications for erosionalandclimaticdriversofweatheringandforcarbon-cyclefeedbacks. Geol- ogy 40:811–814. West AJ, Galy A, Bickle M (2005) Tectonic and Climatic Controls on Silicate Weathering. Earth and Planetary Science Letters 235:211–228. White AF, Brantley SL (2003) The effect of time on the weathering of silicate minerals: why do weathering rates differ in the laboratory and field? Chemical Geology 202:479–506. 152 White AF, Bullen TD, Vivit DV, Schulz MS, Clow DW (1999) The role of dis- seminated calcite in the chemical weathering of granitoid rocks. Geochimica et Cosmochimica Acta 63:1939–1953. White AF, Yee A, Flexser S (1985) Surface oxidation-reduction kinetics associated with experimental basalt-water reaction at 25ÂřC. Chemical Geology 49:73–86. Wigley T, Plummer L, Pearson F (1978) Mass transfer and carbon isotope evolu- tion in natural water systems. Geochimica et Cosmochimica Acta 42:1117–1139. Wildman RA, Hickey LJ, Dickinson MB, Berner RA, Robinson JM, Dietrich M, Essenhigh RH, Wildman CB (2004) Burning of forest materials under late Paleozoic high atmospheric oxygen levels. Geology 32:457–460. Wilhelm S, Maxwell D, Trick C (1996) Growth, Iron Requirements, and Siderophore Production in Iron-Limited Synechococcus PCC 7002. Limnology and oceanography 41:89–97. Willenbring JK, von Blanckenburg F (2010) Long-term stability of global erosion rates and weathering during late-Cenozoic cooling. Nature 465:211–4. Wolff-Boenisch D, Wenau S, Gisalson S, Oelkers EHE, Gislason SR (2011) Dis- solution of basalts and peridotite in seawater, in the presence of ligands, and CO2: Implications for mineral sequestration of carbon dioxide. Geochimica et Cosmochimica Acta 75:5510–5525. Wu N, Farquhar J, Strauss H, Kim ST, Canfield DE (2010) Evaluating the S- isotope fractionation associated with Phanerozoic pyrite burial. Geochimica et Cosmochimica Acta 74:2053–2071. Zhang, J (1995) Carbon isotope fractionation during gas-water exchange and dissolution of CO2. Zhang YG, Pagani M, Liu Z, Bohaty SM, Deconto R (2013) A 40-million-year history of atmospheric CO2. Philosophical Transactions of the Royal Society A 371:1–20. 153 Appendix A Mineral Dissolution Model To mathematically simulate the development and effects of a surface Fe- oxyhydroxide coating during olivine dissolution, I consider a cross-section of a simple "mineral" composed of a 300 by 500 matrix. Here, the x dimension rep- resents the length of the crystal while the y dimension represents depth within the crystal lattice. For each row of the matrix, the cells are populated with a fixed proportion of 1’s and 2’s with the locations of the 2’s within each row being selected using a bounded random integer generator. For the bounds of the random integer generator, I excluded cells within the first and last columns since they are be subject to edge effects and will ultimately be excluded from the analyses (see below). In all of the model simulations, cells with a value of 1 are equivalent to Mg sites and will dissolve if exposed at the surface of the crystal. Once a Mg site is exposed at the surface, the value of the cell is changed from 1 to 0 (where 0 = solution). Cells with a value of 2 are equivalent to Fe sites that, when exposed at the surface, become Fe precipitates and retain a cell value equal to 2. The presence of an Fe precipitate shields the underlying site from contact with the surface and, as a result, can inhibit dissolution. For an Fe site at the surface located in row x and column y, I calculate the sum of the height and connectivity (SHC) of the cell by adding the values (C) of four of the adjacent cells following the equation: SHC (x,y) =C (x+1,y+1) +C (x−1,y+1) +C (x+1,y) +C (x−1,y) (A.1) which, in the formulation, allows SHC values to range from 1 to 8. If the SHC value of an Fe site is greater than 5, then the Fe site persists at the mineral surface until the next time step. If the SHC value of a cell is lower than 5, a random integer value (T) is generated from a defined range that is variable betweendifferentmodelexperiments. IfthethesquareoftheSHCvalueislessthan T, then the Fe precipitate does not inhibit dissolution and cell value is replaced 154 with a zero. If the the square of the SHC value is greater than T, then the Fe site persists at the mineral surface until the next time step. This approach results in different sites having probabilities of inhibiting dissolution proportional to the square of there SHC values. By changing the range of values used to generate T, the probabilities of inhibiting dissolution can be varied between separate model simulations (i.e. the difference between the gray and black curves in Figure 2.5). At each time step, the model starts at the first cell in the first row and queries the value, does any necessary calculations, stores the result in a temporary matrix, and then proceeds to the next cell in the same row. Once the model has queried an entire row, all of the values within that row are updated before the model contin- ues with the subsequent row. This procedure ensures that dissolution propagates downwards and not from left to right. Tocalculatedissolutionrates,thedifferencebetweenthesumofallcellswithina "representativeportion"ofthemineralbeforeandaftereachtimestepiscalculated. For the "representative portion", I remove the first and last 50 columns to avoid edge effects. The dissolution rate at each timestep is then normalized by dividing the value by 100, which is the expected value if all cells dissolved at the same rate. For a each set of model parameters, I report the average and standard deviation of the normalized dissolution rate at each timestep based on 5 independent model runs. 155 Appendix B Feedback Model To model the feedbacks between ligand production and Fe release from miner- als, I consider two coupled ordinary differential equations that describe the time rate of change of free ligand ([S]) concentrations d[S] dt = (P−R− (Q× [S]))/V (B.1) and Fe concentrations ([Fe]) d[Fe] dt = (R−U− (Q× [Fe]))/V (B.2) where P is the [Fe]-dependent ligand production rate, R is the [S]-dependent reac- tion rate of the ligand with an Fe bearing mineral, U is the [Fe]-dependent biotic uptake of the Fe-ligand complex, Q is the flow rate in and out of the system, and V is the total volume of the system. This formulation assumes that dissolved Fe can only be present if it is bound to a ligand. To model the [Fe]-dependent ligand production rate (P), I multiply equation 2.6 by equation 2.7, which yields the equation: P = μ× [Fe] k Fe + [Fe] × −Y max × [Fe] k L + [Fe] +Y max (B.3) where μ max is the maximum growth rate, k Fe is the concentration of Fe where the growth rate is half of its maximum value, k L is the concentration of Fe where siderophore yield (mass of siderophore / biomass) is half its maximum value, and Y max if the maximum siderophore yield. To model the [S]-dependent ligand reaction rate (R), I assume that reaction rates increase with increasing [S] following a Langmuir adsorption isotherm such that R = R max × [S] k S × [S] (B.4) 156 where R max is the maximum reaction rate and k S is the ligand concentration at which reaction rates are one half of their maximum value. To model the [Fe]- dependent uptake rate (U), I multiply equation 2.6 by a constant (φ) that is equal to the concentration of Fe within the biomass. To generate a dimensionless equation, I define [S*] and [Fe*], which are the non-dimensional equivalents of [S] and [Fe] and are equal to: [S ∗ ] = [S]/k S (B.5) and [Fe ∗ ] = [Fe]/k Fe (B.6) respectively. Similarly, I normalize time by the flow rate and system volume such that t ∗ = (t×Q)/V (B.7) where t* is the non-dimensional equivalent to time. Substituting the dimensionless variables into each equation and collecting like terms yields equations: d[S ∗ ] dt ∗ =A −[Fe ∗ ] 2 w + [Fe ∗ ] 1 + [Fe ∗ ] −B [S ∗ ] 1 + [S ∗ ] − [S ∗ ] (B.8) and d[Fe ∗ ] dt ∗ =C [S ∗ ] 1 + [S ∗ ] −D [Fe ∗ ] 1 + [Fe ∗ ] − [Fe ∗ ] (B.9) where A = μ×Y max k s ×Q (B.10) B = R max k s ×Q (B.11) w = (k L /k Fe ) + [Fe ∗ ]× (1 + (k L /k Fe ) + [Fe ∗ ]) (B.12) C = R max k Fe ×Q (B.13) 157 and D = μ×φ k Fe ×Q (B.14) To explore the solution space of equations B.8 and B.9, I first define a reference set of parameter values based on rounded values generated by fitting the Fe release data to equation B.4 and the data of Fekete et al. (1983) to equations 2.6 and 2.7 (Figures 2.3 & 2.6). I then vary L, k L , and Q to observe how physiological parameters (i.e. L and k L ) and environmental conditions (i.e. Q) affect S* values. Steady-state solutions were calculated using the ode45 solver in MATLAB 2013a with initial S* and Fe* values of 10 −3 . 158 Appendix C Atmospheric input correction To estimate the volume-weighted mean of precipitation element/Cl − ratios, the riverdataisassumedtorepresentatwo-componentmixtureofsolutesderivedfrom precipitation and the dissolution of minerals. The solute to chloride concentration ratio in the river (X/Cl − riv ) is thus given by: X/Cl − riv =F precip (X/Cl − precip ) + ((1−F precip )×X/Cl − min ) (C.1) where F precip is the fractional contribution from precipitation, X/Cl − precip is the solute to chloride concentration ratio of precipitation, and X/Cl − min is the solute to chloride concentration ratio that corresponds to mineral weathering. To utilize equationC.1tocalculatethevalueofX/Cl − precip forindividualsolutes, itisassumed that the precipitation volume-weighted Na/Cl − is equal to the Na/Cl − ratio of sea- water (i.e. 0.86). The basis for this assumption is that seawater aerosols represent the major source of Na and Cl − ions to precipitation, the measured Na/Cl − ratios of precipitation are similar to the seawater ratio (Figure 3.2), and previous research suggests that Na and Cl − concentrations are the least affected by biospheric and anthropogenic inputs relative to other elements in Amazonian precipitation (Stal- lard & Edmond, 1981; Andreae et al., 1990a). By setting Na/Cl − precip equal to the Na/Cl − of seawater in equation C.1 written for Na/Cl − riv and arranging the equation to solve for F precip , it can be substituted into equation C.1 written for another solute yielding the linear equation X/Cl − riv =M× (Na/Cl − riv − 0.86) +X/Cl − precip (C.2) where M = ((X/Cl − min - X/Cl − precip ) / (Na/Cl − min - 0.86)). To calculate X/Cl − precip for all solutes other than Na using the relationship described by equation C.2, a linear regression of X/Cl − riv against Na/Cl − riv is evaluated at the seawater Na/Cl − ratio (Figure 3.2). This procedure is performed separately for each element in each catchment, and the uncertainties associated with the fitted parameters of the 159 linear regression are used to determine the uncertainty of the calculated X/Cl − precip . Visual analysis of the relationship between X/Cl − ratios and the Na/Cl − ratios of the river water samples revealed some clear deviations from two component mixing (i.e. multiple lithologic end-members), especially for the samples collected at the Wayqecha gauging station. As a result, the linear regression was only performed with a subset of the samples (the wet-season data collected at high resolution) from the Wayqecha gauging station in order to limit the effect of multiple lithologic sources of solutes on the calculated precipitation end-member. The calculated X/Cl − ratios were used to correct for the atmospheric contribution by multiplying the catchment specific X/Cl − ratio by the measured Cl − concentration in the stream water to determine the amount of the element supplied by atmospheric inputs. This amount is then subtracted from the total amount of the element that was measured in the stream water. The overall correction is described by the equation: X∗ =X riv − (Cl − riv ×X/Cl − precip ) (C.3) where X ∗ is the corrected concentration of an element in the sample, X riv is the measured concentration of an element in the sample, and Cl − riv is the measured concentration of an chloride in the sample. The uncertainty associated with this procedure was determined by propagating the uncertainty in the measured elemen- tal concentrations (RSD = 5%) and the calculated elemental ratios using Gaussian error propagation techniques. For a subset of samples, the uncertainty was also estimated using a Monte Carlo approach assuming that all of the individual uncer- tainties were normally distributed. The two approaches yielded similar values for the 68% confidence interval (i.e. at the 1σ level). Due to the large uncertainty associated with estimating the cyclic salt end-member composition via a linear regression of the data, the uncertainty of the corrected data for a given element shows a strong power-law dependence on the X/Cl − ratio of the sample. 160 Appendix D Detailed methods D.0.1 Solid phase geochemical analyses Samples of river bank sediments and rocks were collected from across the entire study site in order to constrain the elemental and isotopic composition of different lithologic end-members. Sub-samples of the river bank sediments were separated using a riffle splitter and powdered in a ball mill. Rock samples were disaggregated using an agate mortar and pestle before being ground in a ball mill. Bulk XRF measurements To determine the content of non-volatile elements in the river bank sediment samples, the samples were mixed in a 1:2 ratio with lithium metaborate and then doubly fused in graphite crucibles at 1000 ◦ C. The twice fused glass beads were then polished and analyzed by X-ray fluorescence spectroscopy (XRF) at Pomona College. A suite of 35 elements were analyzed. Only measurements of Na, Ca, Mg, and Sr concentrations are discussed in this paper. The stream sediment reference material STSD-2 (Environment Canada) was processed and analyzed using the same procedure in order to check for accuracy. For all reported elements, the measured values of STSD-2 agree with certified values within 10%. Carbonate content and δ 13 C measurements To determine the content and the carbon isotopic composition (δ 13 C V−PDB ) of carbonate in the rock and sediment samples, between 0.1 and 2 grams of the ground powders were weighed into glass exetainers, sealed, and evacuated. To each sample, 4 mL of N 2 -purged 10% H 3 PO 4 was added and allowed to react with the sample powder at room temperature overnight. The evolved CO 2 was stripped from the exetainers using a stream of N 2 using an Automate preparation device and fed into a Picarro G2131-i via a Picarro A0301 Liaison. After the first anal- ysis, all of the exetainers were placed in a heated aluminum block and allowed to 161 react for one week at 60 ◦ C before re-analysis. The measurement of weight percent particulate inorganic carbon (PIC) was calibrated using a solid carbonate stan- dard. The measurement ofδ 13 C was calibrated with in-house carbonate standards that were previously calibrated relative to V-PDB at the University of California- Davis. Additionally, the NBS 18 carbon isotope standard was also run periodically in order to ensure accuracy. Solid phase Sulfur isotope measurements To measure the sulfur isotopic composition (δ 34 S V−CDT ) of reduced sulfur min- erals (i.e. S 50 ) present within rock samples, the reduced sulfur compounds were converted toH 2 S and precipitated asAg 2 S using the chromium reduction method of Gröger et al. (2009). Briefly, the sample powders were mixed with ethanol and concentrated HCl and then reacted with an acidicCr 2+ Cl 2 solution in aN 2 -flushed digestion vessel. During the reaction, the digestion vessels were heated from below with a hotplate. The liberatedH 2 S gas was passed through a condenser and bub- bled through a solution ofAgNO 3 andNH 4 OH in order to trap S 2− as Ag 2 S. For each sample, approximately 1 gram of powder was reacted for one hour. After the reaction was completed, the Ag 2 S was separated from the AgNO 3 and NH 4 OH solution by centrifugation, rinsed three times with de-ionized water (DIW; 18.2 MΩ resistivity), and dried overnight in an oven at 60 ◦ C. Sub-samples of the homogenized Ag 2 S powders were sent to the University of Arizona stable isotope lab and the sulfur isotopic composition was measured by IRMS. To check the accuracy and reproducibility of the chromium reduction procedure, an in-house pyritestandardwasprocessedduringeachsessionandtheprocedurewasreplicated for select samples. Overall, calculated yields for the samples and standards were similar (80-90 %) and the isotopic composition of the Ag 2 S produced from the pyrite standard was identical, within the analytical uncertainty (0.15h), to the un-processed pyrite. Sequential river bank sediment leaches In order to selectively dissolve carbonate minerals in the river bank sediment samples, a sequential leaching procedure based on the method of Leleyter & Probst (1999) was used. Both ground and un-ground sediment samples were tested, but 162 only the leaches of the un-ground sediment samples were found to selectively dis- solve carbonates to a degree that allowed for the determination of their chemistry. For the un-ground samples, 4-6 grams of sediment were separated from the total sample using a riffle splitter and then split into two roughly equal aliquots that were leached separately in 50 mL polypropylene (PP) centrifuge tubes. Duringeachleachingstep, thesedimentsampleswerekeptatroomtemperature and stirred by laying the tubes on a shaking table set to 200 rpm. After each leachingstep, theleachatewasfirstseparatedfromthesedimentsbycentrifugation. The supernatant was then decanted and filtered with a 0.2μm nylon filter. Before the next leaching step, the sediment samples were rinsed three times with de- ionized water. To remove soluble salts, the sediments were first leached with 10 mL of DIW for 30 minutes. Next, exchangeable elements were removed by leaching the sediments with 10 mL of 1M NH 4 Cl for 2 hours. Finally, carbonate minerals were selectively dissolved by leaching the sediments for 5 hours with 10 mL of a 1M acetic acid solution that was set to a pH of∼4.5 by titration with NH 4 OH. After filtration, the acetic acid leachates were evaporated to dryness in PP vials in a clean laboratory and then re-dissolved in 5% HNO 3 . The concentrations of Al, Si,Na,Ca,Mg,Sr,andLiintheleachatesolutionsweredeterminedwithanAgilent 4100 microwave plasma atomic emission spectrometer (MP-AES) calibrated using synthetic standards. The results are reported as nanomoles of element leached per gram of sample. All of the acetic acid leachate solutions contain Al and Si, which suggests that both carbonate and silicate minerals were dissolved during the leaching procedure (FigureD.1). Tocorrectforthis, asimplemixingmodelwasdeveloped. Thismodel assumesthatcarbonatesmineralscontainnoAl, whichallowsforthefractionofCa sourced from silicate mineral dissolution (f Si ) to be calculated with the equation: f Si = Al/Ca measured Al/Ca silicate (D.1) where Al/Ca measured is the measured Al/Ca ratio of the leachate and Al/Ca silicate is the Al/Ca ratio of silicate minerals. Based on XRF measurements of the bulk sediment samples with low carbonate contents, the Al/Ca silicate ratio is assumed to be 100 mol/mol for the mixing calculation (Figure D.1). 163 With the results of equation D.1, an element/Ca ratio of the carbonate end- member (X/Ca carb ) can be calculated using the equation: X/Ca carb = X/Ca measured − (X/Ca silicate ×f sil ) 1−f sil (D.2) where X/Ca measured is the element/Ca ratio measured in the leachate solution and X/Ca silicate is the element/Ca ratio of the silicate end-member. Based on XRF measurements of the bulk sediment samples with low carbonate contents, the Mg/Ca silicate , Sr/Ca silicate , and Na/Ca silicate ratios are assumed to range from 0.1-15 mol/mol, 30-60 mmol/mol, and 1-5 mol/mol respectively (Figure D.1). To account for the observed ranges in the X/Ca silicate values, the mixing cal- culations were performed using 100 evenly spaced values of X/Ca silicate that span the entire observed range. The results were then filtered to separate results where the minimum calculated X/Ca carb was less than zero. For samples where all of the calculated X/Ca carb values were greater than 0, the 16 th , 50 th and 86 th percentiles are reported. For samples where some of the calculated X/Ca carb values were less than 0, only the 50 th and 86 th percentiles are reported (since these data points do not constrain the lower bound of X/Ca carb ). D.0.2 Dissolved phase geochemical analyses Water samples were collected using slightly different methods depending upon the sampling year. For samples collected before 2012, water was collected from the river surface using a clean PP bottle, filtered onsite with a 0.2 μm nylon filter, and split into two 60 mL high-density polyethylene bottles (HDPE) and one glass exetainer. One of the 60 mL HDPE bottles was preserved with 2 drops of high purity HCl dispensed from an acid-washed Teflon dropper bottle for cation analyses. The other HDPE bottle and the glass exetainer were left unpreserved. In the laboratory, samples with any remaining particulates (e.g., from flocculated aggregates forming after field filtration) were re-filtered before analysis with a 0.2 μm nylon filter. After 2012, water samples were collected from the river surface with a clean PP bucket and transferred to 10 L plastic bags before filtration. Within 24 hours of collection, the samples were filtered with 0.2 μm polyethersulfone (PES) filters 164 housed in an teflon filtration unit with a peristaltic pump and tygon tubing. The filtrate was collected directly into two clean 60 mL HDPE bottles and one glass exetainer. One of the 60 mL HDPE bottle was preserved with 60 μL of concen- trated distilled HNO 3 dispensed from a teflon vial with an acid-washed pipette tip. The other HDPE bottle and the glass exetainer were left unpreserved. Cation and Si concentrations To determine the concentrations of Na, K, Ca, Mg, Si, Li, and Sr, the acid- ified water samples were analyzed using an MP-AES calibrated with synthetic standards. Precision and accuracy was assessed by analyzing a reference material every 15 samples. For Ca, Mg, Na, K, and Si, the reference material ION-915 was used (Environment Canada). For Li, the reference material TMDA-51.4 (Environ- ment Canada) was used. For Sr, an in-house prepared SrCO 3 solution was used. Replicate analyses of each solution reveals an analytical precision within 5% (1σ) for each analyte. Anion concentrations To determine the concentrations of Cl − and SO 2− 4 , the un-acidified samples were analyzed with a Metrohm ion chromatograph equipped with a Metrosep A4/150 column and a conductivity suppressor. The elements were eluted from the column with 3.2 mM Na 2 CO 3 and 1.0 mM NaHCO 3 at a flow rate of 0.7 mL min −1 . The instrument was calibrated using synthetic standards. Precision and accuracy was assessed by analyzing a certified reference material (ION-915, Envi- ronment Canada) after every 15 samples. Replicate analyses of ION-915 reveals an analytical precision within 5% (1σ) for each analyte. DIC concentration and δ 13 C measurements To measure the concentration and carbon isotopic composition of dissolved inorganic carbon (DIC), between 7 and 8 mL of sample that had been stored without any headspace in glass exetainers was transferred to an evacuated exe- tainer and acidified with concentrated H 3 PO 4 . The evolved CO 2 was stripped from the exetainers using a stream of N 2 using an Automate preparation device and fed into a Picarro G2131-i via a Picarro A0301 Liaison. The instrument was 165 calibrated identically to what is reported above for the solid carbonate concentra- tion and carbon isotope measurements (see section D). Comparisons between the solid carbonate standard and a seawater reference material (Dickson) revealed a consistent and small (i.e. a few percent) underestimation of DIC in the seawater reference due to the incomplete removal of dissolved CO 2 from the seawater by flushing with N 2 gas. I do not correct for this effect since it negligible relative to the overall reproducibility of the DIC concentration and isotopic measurements. Sulfate-sulfur isotope measurements To measure the δ 34 S V−CDT of dissolved SO 2− 4 , SO 2− 4 was purified from∼ 1-10 mL of sample using either a cation or anion exchange resin following established protocols (Paris et al., 2013). Before separation, all samples were evaporated to dryness within a clean laboratory. For samples purified using a cation exchange resin, the sample residue was re-dissolved in 0.25 % HCl and introduced into a column containing Bio-Rad AG50X8 resin following Paris et al. (2013). For sample purified using an anion exchange resin, the sample residue was re-dissolved in 0.5 % HCl and introduced into a column of AG1X8 resin following Paris et al. (2014). After elution from the columns, the samples were evaporated to dryness and then re-dissolved in 5% HNO 3 . Before analysis, all samples were diluted and mixed with a sodium solution to match the sodium and sulfate concentrations of the bracketing standard. The samples were then analyzed using a Thermo Neptune Plus multi-collector inductively coupled plasma mass spectrometer (MC-ICP-MS) at Caltech using sample-standard bracketing to correct for instrumental drift and mass bias following Paris et al. (2013). Strontium isotope measurements To measure the radiogenic ( 87 Sr/ 86 Sr) and stable (δ 88 Sr) isotopic composition of dissolved Sr, the acidified samples were purified using an automated HPLC separation with Sr Spec resin at the Institute de Physique du Globe Paris (IPGP). The purified samples were evaporated to dryness, re-dissolved in 0.5 M HNO 3 and analyzed on a Thermo Neptune plus MC-ICP-MS at IPGP. Sample-standard bracketing was preformed using the SRM987 standard to correct for instrumental massbias. TocorrectforKrinterferences, the 83 Kr/ 84 Krand 83 Kr/ 86 Krratioswere 166 determined using the blank solution at the beginning of the run and the 83 Kr signal was monitored for each sample and standard. To correct for Rb interferences, the 87 Rb signal of a 5 ppb Rb solution was measured at the beginning of the run. Water isotope measurements The δ 18 O and δD composition of water samples were measured using a Los Gatos DLT-1000 liquid water isotope analyzer as described in Ponton et al. (2014). For each sample, the mean and standard deviation of eight replicate injections into the instrument are reported. The instrument was calibrated relative to the VSMOW scale with two working standards. Laboratory and method inter- comparison between two laser instruments (Los Gatos and Picarro) and an IRMS revealed no biases or artifacts for this sample set (Clark et al., 2014). 167 10 −1 10 0 10 1 10 2 10 3 10 −2 10 −1 10 0 10 1 10 2 Al/Ca (mol/mol) Mg/Ca (mol/mol) 10 −1 10 0 10 1 10 2 10 3 10 0 10 1 10 2 Al/Ca (mol/mol) Sr/Ca (mmol/mol) 0 5 10 15 20 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Al/Ca (mol/mol) Mg/Ca (mol/mol) 0 5 10 15 20 0 5 10 15 Al/Ca (mol/mol) Sr/Ca (mmol/mol) Acetic acid leachates Riverbank sediments (low carbonate) silicate end-member used for mixing Acetic acid leachates Silicate-corrected Mg/Ca (group 1) Silicate-corrected Mg/Ca (group 2) Figure D.1: Correction for silicate contamination of leachates. (A) Vari- ation in the Mg/Ca ratio of the riverbank leachate samples (circles) with the Al/Ca ratio. (B) ariation in the Sr/Ca ratio of the riverbank leachate samples (circles) with the Al/Ca ratio. The elemental ratios of the riverbank samples with low carbonate contents (crosses) are shown for reference. The gray region defines the silicate end-member used for the correction based on equation D.2. (C) Raw leachates(circles)andsilicate-correctedcarbonateend-member(diamonds)Mg/Ca ratios. (D) Raw leachates (circles) and silicate-corrected carbonate end-member (diamonds) Sr/Ca ratios. 168 Appendix E Tributary mixing model I used the water isotope composition (δD) and dissolved chloride concentration as conservative tracers for mixing calculations. In order to calculate the relative contributions of the Alto Madre de Dios, Manu, Blanco, Chirbi, and Colorado riverstothedischargeoftheMadredeDiosriverattheCICRAgaugingstation,the difference in tracer composition (δD or Cl) between the mainstem upstream of the confluenceandthetributaryneedstoberesolvableabovetheanalyticaluncertainty of the tracer measurement (i.e. >3σ). All of the mixing pairs except for the Madre de Dios - Colorado during the wet season showed significant differences in δD values. NoneofthemixingpairsexceptfortheMadredeDios-Coloradoduringthe wet season showed significant differences in Cl concentrations. Consequently, all of the mixing calculations were performed usingδD except for the mixing between the Madre de Dios and Colorado rivers during the wet season, which uses Cl concentrations. For the tributary mixing model, I use a two end-member mixing equation: Tracer MS−down = (f trib ×Tracer trib ) + ((1−f trib )×Tracer MS−up ) (E.1) where Tracer MS−down is the concentration (Cl − ) or ratio (δD) of the tracer in the mainstem downstream of the confluence, f trib is the fractional contribution of the tributary to the total water flux downstream of the confluence, Tracer trib is the mass of the tracer in the tributary, and Tracer MS−up is the mass of tracer in the mainstem upstream of the confluence. I account for analytical uncertainty by randomly sampling from a normal distribution for each of the measured values. For each confluence, 1000 random samples of δD values or Cl concentrations are used. If a calculation produces a negative values, it is converted to 0. To estimate the absolute water fluxes (i.e. water discharge; m 3 /sec) of each tributary from the relative contributions calculated with equation E.1, I use the the total discharge measured at CICRA using an acoustic doppler current profiler 169 (ADCP) at the time of sampling as a constraint (West, unpublished data). Since the total discharge at CICRA is known, the total discharge of the mainstem and the tributary that join immediately upstream of CICRA (the Colorado river) can be calculated by multiplying their relative contributions calculated with equation E.1 by the total discharge measured using the ADCP. This approach can then be repeated for each upstream confluence after subtracting the calculated discharge of each downstream reach from the total. 170 Appendix F Lithologic source mixing model With five end-members and the assumption that all of the mixing fractions sum to one, four tracers are needed for the inversion component of the mixing model. Here, I use Cl/Ca, Na/Ca, Mg/Ca and Sr/Ca ratios as tracers. The procedure for defining the elemental ratio composition of each end-member, which is a required input for the inversion calculation, is described in detail below. For the elemental ratio composition of the "shale" and "granite" silicate end- members, I define a uniform distribution of Na/Ca and Sr/Ca ratios with a range equal to what is observed in the bulk river bank sediments. For the low Mg/Ca silicate end-member (i.e granite), a uniform distribution with range of 0.1-0.4 is used based on a subset of the bulk river bank sediments. For the high Mg/Ca silicate end-member (i.e. shale), a uniform distribution with range of 1-20 is used based on a subset of the bulk river bank sediments. For both the shale and granite end-members, Cl/Ca values are set equal to zero. For the Mg/Ca and Sr/Ca ratio composition of the Ca-carbonate end-member, I use uniform distributions with ranges defined by the measurements of the river bank sediment leachates after correction for silicate contamination. Additionally, I exclude the samples with Mg/Ca ratios greater than 0.3 since these samples are likely influenced by dolomite dissolution. For the dolomite end-member, I use the same range of Sr/Ca values as the Ca-carbonate end-member, but allow Mg/Ca ratios to vary between 0.8 and 1.2 mol/mol. For both the Ca-carbonate and dolomite end-members, Na/Ca and Cl/Ca values are set equal to zero. Foratmosphericdepositionend-member, IusetheresultsofTorresetal. (2015) where elemental ratios of the atmospheric end-member were calculated by fitting the river data with a mixing equation. Here, I use a uniform distribution of each elemental ratio with a range equal to the 95% confidence interval of the values calculated by Torres et al. (2015). Specifically, the results calculated for the Alto Madre de Dios river at MLC catchment are used since they encompass the results from each of the other catchments. 171 To perform the inversion, a single value for each end-member is drawn ran- domly from the appropriate uniform distribution and used to calculate the mixing proportions of a sample with the mldivide command in MATLAB 2013a. For each sample, this pseudo-random sampling of different end-member ratios is repeated 10 4 times in order to ensure that different combinations of end-member ratios are used. The specific number of calculations was chosen in order to have the 16 th , 50 th , and 84 th percentiles of the distribution of calculated mixing proportions not vary by more than 5% for replicate calculations of a single sample using different sets of randomly generated end-member ratios. If one of the 10 4 sets of results includes a mixing proportion less than zero, that entire set is removed from the final distribution of mixing proportions. Of the 10 4 calculations, the number that produce realistic mixing proportions (i.e. all proportions greater than 0) ranges from 35-2531. Additionally, the set of end- member ratios that corresponds with each set of realistic mixing proportions is logged for each sample. For each calculation that produces a set of realistic mixing proportions, the 87 Sr/ 86 Sr and Li/Ca ratios of the sample are calculated. For 87 Sr/ 86 Sr, the sample value ( 87 Sr/ 86 Sr modeled ) is calculated with the equation: 87 Sr/ 86 Sr modeled = Σ(fCa i ×Sr/Ca i × 87 Sr/ 86 Sr i ) (F.1) where fCa i is the fraction of Ca from the ith end-member calculated from the inversion model, Sr/Ca i is the Sr/Ca ratio of the ith end-member used in the specificinversioncalculation, and 87 Sr/ 86 Sr i isthestrontiumisotopecompositionof theithend-member. Forthe 87 Sr/ 86 Srratiosofthegraniteandshaleend-members, a slightly expanded range (i.e.± 0.01) based on values from Dellinger et al. (2015) is used. For the Ca-carbonate and rain end-members, values from Gaillardet et al. (1997) are used. Since there are no available constraints on the 87 Sr/ 86 Sr ratio of the dolomite end-member, I set it equal to either the Ca-carbonate or shale end-member (see discussion in section 4.5.2). Similarly, Li/Ca ratios are calculated with the equation: Li/Ca modeled = Σ(fCa i ×Na/Ca i ×Li/Na i ) (F.2) 172 where fCa i is the fraction of Ca from the ith end-member calculated from the inversion model, Na/Ca i is the Na/Ca ratio of the ith end-member used in the specific inversion calculation, and Li/Na i is the Li/Na ratio composition of the ith end-member. For the Li/Na ratios of the granite and shale end-members, values from Dellinger et al. (2014) are used. For the Ca-carbonate, dolomite, and rain end-members, Li/Na ratios are assumed to be equal to 0. 173

Abstract (if available)

Abstract When exposed at Earth’s surface, rocks are out of thermodynamic equilibrium with respect to their environment. This disequilibrium drives the chemical transformation, or weathering, of these rocks into soils and governs the chemical composition of natural waters and the atmosphere. Over geologic timescales, complex feedbacks associated with weathering processes are presumed to regulate the concentrations of CO₂ and O₂ in the atmosphere with profound implications for the habitability of the planet. However, a mechanistic understanding of how biologic, tectonic, and climatic conditions interact to control weathering fluxes has remained elusive. In part, our understanding of weathering processes is hindered by the fact that they operate continuously over an enormous range of spatial (atomic to global) and temporal (microseconds to millions of years) scales, but we can only make measurements over discrete ranges of these values. While each scale of observation available offers unique insights, it is often difficult to link observations made at different scales. For my Ph.D., I focused on three distinct projects that span the range of observable scales in order to better understand the links between chemical weathering and long-term biogeochemical cycles. ❧ Chapter 2: Laboratory insights into microbial mineral dissolution. Rocks and minerals represent a major reservoir of bio-essential nutrients. While abundant, some of these lithogenic nutrients, like iron, are not readily bio-available. As a result, many organisms produce metal-binding ligands to scavenge these trace nutrients from the environment. Using targeted laboratory experiments with live microbial cultures and purified microbial ligands, I explored efficacy by which microbes can access trace nutrients from common silicate minerals (Torres et al., in prep). In addition to providing insight into biological nutrient scavenging strategies, this work also provides the basic research necessary to develop microbe-based CO₂ sequestration techniques since the dissolution of silicate minerals for nutrient acquisition also sequesters CO₂. ❧ Chapters 3 & 4: Geomorphic control on the hydrology and carbon budget of weathering. Erosional processes and hydrology are known to influence chemical weathering rates by controlling the timescales over which minerals react. Accurately describing the complex linkages between weathering, erosion, and hydrology observed in natural environments remains a major research challenge. To help address this problem, a major part of my Ph.D. was focused on characterizing how chemical weathering and hydrology are coupled in distinct erosional environments. This work combines hydrologic monitoring, solute chemistry, and water isotope analyses in order to robustly document how water is stored in catchments and link this to measured solute fluxes from chemical weathering (Clark et al., 2014

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Environmental controls on alkalinity generation from mineral dissolution: from the mineral surface to the global ocean

PDF

Ages, origins and biogeochemical role of water across a tropical mountain to floodplain transition

PDF

Germanium and silicon isotope geochemistry in terrestrial and marine low-temperature environments

PDF

Lithium isotopes as a carbon cycle proxy: examining the effects of high-temperature weathering, understanding brachiopod archives, and validating Cenozoic records

PDF

Exploring temporal changes in surface species on weathered feldspar mineral surfaces using solid-state NMR spectroscopy

PDF

Exploring bacteria-mineral interactions

PDF

Critical zone response to perturbation: from mountain building to wildfire

PDF

Habitability of Saturn's moon Titan for acetylenotrophy: laboratory culturing and energetics as tools for the search for life

PDF

Paleoenvironments and the Precambrian-Cambrian transition in the southern Great Basin: Implications for microbial mat development and the Cambrian radiation

PDF

Marine bacterioplankton biogeography over short to medium spatio-temporal scales

PDF

Concentration and size partitioning of trace metals in surface waters of the global ocean and storm runoff

PDF

Investigations in the field of geobiology through the use of digital holographic microscopy

PDF

Evaluation of the budget for silicic acid in Cascadia Basin deep water

PDF

Calcite and aragonite dissolution in seawater: kinetics, mechanisms, and fluxes in the North Pacific

PDF

Earthquake-driven landsliding, erosion and mountain building: from the eastern Tibetan mountains towards global models

PDF

Recent variability in the hyrdological cycle of tropical Asia from oxygen isotopes of tree celulose

PDF

Microbe to microbe: monthly microbial community dynamics and interactions at the San Pedro Ocean Time-series

PDF

Investigating the global ocean biogeochemical cycling of alkalinity, barium, and copper using data-constrained inverse models

PDF

Investigating microbial biofilm community mediated processes on surfaces: from single cell genomics to community meta-omics

PDF

Multi-scale damage signatures across major strike-slip faults

Asset Metadata

Creator Torres, Mark Albert (author)

Core Title Chemical weathering across spatial and temporal scales: from laboratory experiments to global models

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Geological Sciences

Publication Date 07/30/2015

Defense Date 05/26/2015

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag OAI-PMH Harvest,olivine,pyrite,rivers,silicate,weathering

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor West, A. Joshua (committee chair), Amend, Jan (committee member), Fischer, Woodward (committee member)

Creator Email markt253@gmail.com,marktorr@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-616350

Unique identifier UC11303117

Identifier etd-TorresMark-3763.pdf (filename),usctheses-c3-616350 (legacy record id)

Legacy Identifier etd-TorresMark-3763.pdf

Dmrecord 616350

Document Type Dissertation

Format application/pdf (imt)

Rights Torres, Mark Albert

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA