Predictive models for strontium isotope distributions in bedrock, water and environmental materials for regional provenance studies

Strontium isotope ratio (87Sr/86Sr) has a strong potential to complement atmospherically-derived traditional stable isotopes in geochemical provenance studies because strontium (Sr) in Earth surface reservoirs is sourced from local bedrock. As such, 87Sr/86Sr variations are discrete and differ drastically from the large scale smoothed variations of atmospherically-derived stable isotopes. Among the most successful recent applications, 87Sr/86Sr has been used to interpret provenance of individuals in archeology, to identify the origin of dust aerosols, to reconstruct cation source and mobility in rivers, and to reconstruct animal or material movement pathways. However, extending the applications of 87Sr/86Sr for provenance to larger spatial scales is currently hampered by the absence of methods to predict the 87Sr/86Sr of Sr sources at the regional scale. In this dissertation, a flexible geostatistical framework is established to predict 87Sr/86Sr distributions in bedrock, river water and soil water at regional scale. This approach leverages publically-available geospatial data on rock geochemistry, surficial and bedrock geology, climate, hydrology, and aerosols to model the input and propagation of Sr from multiple geological sources through hydrosystems and ecosystems. In a first step, we develop predictive models for 87Sr/86Sr in bedrock as a function of variations in rock age and rock type. In a second step, we model the Sr release from different rock units, its transport as dissolved Sr or in aerosols, and its accumulation and mixing in ecosystems. The model was tested for the contiguous USA and circum-Caribbean region and the model showed promising results but the predictive power remained too low for routine provenance interpretations. In a final step, we develop a flexible geochemical framework that explicitly accounts for prediction uncertainty and local variability of 87Sr/ 86Sr and includes a Sr-specific process-based chemical weathering model. This improved model version is applied to predict 87Sr/86Sr in bedrock and rivers over Alaska and explain 82% of 87Sr/ 86Sr variance in Alaska Rivers. Integrated into a multi-isotopes framework, 87Sr/86Sr could dramatically improve the spatial resolution of provenance assignments. Predictive 87Sr/86Sr models are also a powerful standalone tool to visualize, identify and model mechanistic processes influencing local to global 87Sr/86Sr in Earth surface reservoirs.

PREDICTIVE MODELS FOR STRONTIUM ISOTOPE DISTRIBUTIONS IN BEDROCK, WATER AND ENVIRONMENTAL MATERIALS FOR REGIONAL PROVENANCE STUDIES by Clement Pierre Bataille A dissertation submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Geology Department of Geology and Geophysics The University of Utah December 2014 Copyright © Clement Pierre Bataille 2014 All Rights Reserved The Unive r s i ty of Utah Graduat e School STATEMENT OF DISSERTATION APPROVAL The dissertation of Clement Pierre Bataille has been approved by the following supervisory committee members: Gabriel J. Bowen , Chair 06/02/14 Date Approvea Diego P. Fernandez , Member 06/02/14 Date Approved Lisa Stright , Member 06/02/14 Date Approved Thure E. Cerling , Member 06/02/14 Date Approved Brett J. Tipple , Member 06/17/14 Date Approved and by John Bartley , Chair/Dean of the Department/College/School of Geology and Geophysics and by David B. Kieda, Dean of The Graduate School. ABSTRACT Strontium isotope ratio (87Sr/86Sr) has a strong potential to complement atmospherically-derived traditional stable isotopes in geochemical provenance studies because strontium (Sr) in Earth surface reservoirs is sourced from local bedrock. As such, 87Sr/86Sr variations are discrete and differ drastically from the large scale smoothed variations of atmospherically-derived stable isotopes. Among the most successful recent applications, 87Sr/86Sr has been used to interpret provenance of individuals in archeology, to identify the origin of dust aerosols, to reconstruct cation source and mobility in rivers, and to reconstruct animal or material movement pathways. However, extending the applications of 87Sr/86Sr for provenance to larger spatial scales is currently hampered by the absence of methods to predict the 87Sr/86Sr of Sr sources at the regional scale. In this dissertation, a flexible geostatistical framework is established to predict 87Sr/86Sr distributions in bedrock, river water and soil water at regional scale. This approach leverages publically-available geospatial data on rock geochemistry, surficial and bedrock geology, climate, hydrology, and aerosols to model the input and propagation of Sr from multiple geological sources through hydrosystems and ecosystems. In a first step, we develop predictive models for 87Sr/86Sr in bedrock as a function of variations in rock age and rock type. In a second step, we model the Sr release from different rock units, its transport as dissolved Sr or in aerosols, and its accumulation and mixing in ecosystems. The model was tested for the contiguous USA and circum-Caribbean region and the model showed promising results but the predictive power remained too low for routine provenance interpretations. In a final step, we develop a flexible geochemical framework that explicitly accounts for prediction uncertainty and local variability of 87Sr/ 86Sr and includes a Sr-specific process-based chemical weathering model. This improved model version is applied to predict 87Sr/86Sr in bedrock and rivers over Alaska and explain 82% of 87Sr/ 86Sr variance in Alaska Rivers. Integrated into a multi-isotopes framework, 87Sr/ 86Sr could dramatically improve the spatial resolution of provenance assignments. Predictive 87Sr/86Sr models are also a powerful standalone tool to visualize, identify and model mechanistic processes influencing local to global 87Sr/86Sr in Earth surface reservoirs. iv To my parents and my wife for their loving support throughout the years, to anonymous French and American taxpayers who generously supported my education, to Yves Godderis and Emmanuel Brochier for transmitting to me their passion of natural sciences and metaphysics, and to God for creating strontium isotopes and making the Universe so fascinating. TABLE OF CONTENTS ABSTRACT................................................................................................................................ iii LIST OF TABLES......................................................................................................................viii LIST OF FIGURES......................................................................................................................x ACKNOWLEDGMENTS.........................................................................................................xii CHAPTER I. INTRODUCTION................................................................................................................... 1 Isotope geochemistry in provenance studies................................................................... 2 Strontium isotope geochemistry for provenance studies................................................4 Sr isotope systematics..........................................................................................................5 Strontium cycle.................................................................................................................... 6 87 86 Sr/ Sr variations in the Earth throughout geological tim e .........................................7 Application of8 7 Sr/8 6 Sr in low temperature geochemistry............................................ 8 Modeling 87Sr/86Sr variations...........................................................................................11 Objectives and outline...................................................................................................... 13 References .......................................................................................................................... 14 II. MAPPING 87Sr/86Sr VARIATIONS IN BEDROCK AND WATER FOR LARGE SCALE PROVENANCE STUDIES....................................................................................... 21 Abstract..............................................................................................................................22 Introduction.......................................................................................................................22 Bedrock models.................................................................................................................23 Water models......................................................................................................................25 Results and discussion...................................................................................................... 26 Conclusion and perspectives............................................................................................32 Acknowledgements ........................................................................................................... 33 References .......................................................................................................................... 33 Supplementary methods and tab le s................................................................................ 36 III. MAPPING MULTIPLE SOURCE EFFECTS ON THE STRONTIUM ISOTOPIC SIGNATURES OF ECOSYSTEMS FROM THE CIRCUM-CARRIBBEAN REGION ....................................................................................................................................85 Abstract..............................................................................................................................86 Introduction.......................................................................................................................86 Material and methods........................................................................................................88 Discussion.......................................................................................................................100 Acknowledgements.........................................................................................................102 Literature cited.................................................................................................................102 Appendix A .................................................................................................................... 106 IV. A GEOSTATISTICAL FRAMEWORK TO PREDICT STRONTIUM ISOTOPES VARIATIONS IN ALASKA RIVERS................................................................................ 110 Abstract ............................................................................................................................ 111 Introduction.....................................................................................................................112 Material and methods..................................................................................................... 115 Results and discussion....................................................................................................132 Conclusion.......................................................................................................................143 Acknowledgements.........................................................................................................144 References.......................................................................................................................144 Supplementary material..................................................................................................165 Supplementary dataset 1: Bedrock model validation dataset....................................203 Supplementary dataset 2: Chemical weathering model calibration dataset........... 204 References.......................................................................................................................205 V. CONCLUSION................................................................................................................... 207 Summary .......................................................................................................................... 208 Perspectives.....................................................................................................................209 Limitations and future improvements.......................................................................... 211 References.......................................................................................................................214 vii LIST OF TABLES 2.1 Geology and measured and modeled 87Sr/86Sr values for bedrock in the catchment water model validation catchments..........................................................................................29 25.1 Values of (Rb/Sr)parent and (Rb/Sr)iithoiogy used in equations 2 and 3 for each unique lithologic descriptor (rocktypel and rocktype2) present in the USGS geodatabases (U.S. Geological Survey, 2005, State Geologic Map Compilation, online at http://tin.er.usgs.gov/geology/state/) ....................................................................................... 44 25.2 Assigned numerical age for each unique age descriptor present in the composite geodatabase.................................................................................................................................49 25.3 Estimated values of f 7S r / 6Sr)seawater throughout Earth history as used in equation 3 ....................................................................................................................................75 25.4 Values of Sr content and W used in equations 5 and 6 for each unique lithologic descriptor (rocktypel and rocktype2) present in the composite geodatabase....................79 25.5 Bulk dissolution rates of common minerals in laboratory as found by Franke (2009).............................................................................................................................84 3.1 Chemical weathering model validation............................................................................94 3.2 Mixing model validation.................................................................................................... 95 3.A1 Parameterization of Eqs. 5 and 7 for each lithological descriptor present in the Caribbean geodatabase (French et al. 2004)........................................................................ 106 4.1 Summary statistics of the chemical weathering model calibration steps.................. 152 4.2 Summary of goodness of fit measures for the bedrock model over Alaska..............153 4.3 Summary of goodness of fit measures for the chemical weathering model for Alaska River......................................................................................................................................... 154 4.4 Summary of goodness of fit measures for the catchment water model.....................155 4.5 Summary of lithological proportion across Alaska and their corresponding mean slope and mean permafrost zonation index (pfi)................................................................. 156 4.51 Summary of power-transformation for f 7R b /6Sr)parent dataset of each major GLiM lithological class.......................................................................................................................174 4.52 Q-Q plots and histograms of (87R b /6Sr)parent dataset of each major GLiM lithological class after power-transformation and outliers removal..................................175 4.53 Summary of power-transformation for f 7R b /6Sr)rock for each major GLiM lithological class.......................................................................................................................177 4.54 Q-Q plots and histograms of (87R b /6Sr)rc^ck datasets for each major GLiM lithological class after power-transformation.......................................................................178 4.55 Summary of variogram model proprieties used in ordinary krigging of power-transformed (87R b /6Sr)rock of each GLiM class dataset.....................................................181 4.56 Plot of empirical variograms and fitted variogram models for each GLiM major lithological class.......................................................................................................................182 4.57 Scatterplot of predicted vs. observed power-transformed (87Rb/86Sr)rock for each lithological class and summary statistic of model prediction............................................ 185 4.58 Attribute table of the digital Atlas of Terranes from the Northern Cordillera (Colpron and Nelson, 2010) with additional lithological description, min age and max age field based on Nelson et al., 2 0 1 3 ..........................................................................................189 ix LIST OF FIGURES 2.1 Three-stage model for the evolution of the 87Sr/86Sr in Earth materials through geological tim e ........................................................................................................................... 23 2.2 Location of the samples used for the water model verification: Green: 87Sr/86Sr measurement in stream water from 1: Clarks Fork of the Yellowstone Basin (Horton et al., 1999); 2: Owens River Lake Basin (Pretti and Stewart, 2002); 3: Scioto River watershed (Stueber et al., 1975); 4: Susquehanna River Basin (Fisher and Stueber, 1976); 87 86 Red: Sr/ Sr measurements of marijuana from 79 counties across the USA (West et al., 2009)............................................................................................................................................27 2.3 Validation of silicate Sr isotope model.............................................................................27 2.4 Validation of carbonate Sr isotope model, showing linear correlations between 246 87 86 87 86 worldwide Sr/ Sr measurements for carbonate rocks and the Sr/ Sr predicted by the bedrock model and age-only models....................................................................................... 28 2.5 Validation of silicate Sr isotope model by lithology.......................................................28 2.6 Modeled bedrock Sr isotope ratios for the contiguous U SA .........................................28 2.7 Catchment water model validation results.......................................................................29 2.8 Validation of the catchment water Sr isotope model across all study catchments, showing linear regressions between measured 87Sr/86Sr and flux-weighted catchment water and age-only water model predictions for 68 streams of the USA (celestite-corrected values are used for the Scioto River)......................................................................31 2.9 Modeled Sr isotope ratios for (A) local water, (B) flux-weighted catchment water, (C) flux-weighted catchment water model values averaged within watersheds of the Watershed Boundary Dataset (http://www.nrcs.usda.gov/wps/portal/nrcs/main/national/water/watershed/dataset) and (D) residual between major bedrock (Fig. 6B) and average flux-weighted catchment water (Fig. 9C )........................................................................................................................... 32 2.10 Linear regression between 87Sr/86Sr in marijuana and the mean values of the modeled water in the country of sample origin......................................................................33 3.1 Representation of the box model with fluxes of dissolved Sr mixing in the soluble bioavailable Sr pool................................................................................................................... 88 3.2 Bedrock-only model 87Sr/86Sr variations calculated using equations from Bataille and Bowen (2012).............................................................................................................................91 3.3 Sample locations for data included in the validation datasets....................................... 96 3.4 Bioavailable Sr isotope model validation results............................................................ 97 3.5 (A) Contribution of bedrock weathering to the bioavailable Sr pool calculated as ^(Sr)w^bio/(F(Sr)w^bio+F(Sr)ss^bio+F(Sr)d^bio).......................................................................98 3.6 (A) Difference between predicted 87Sr/86Sr from the bedrock-only model and 87Sr/86Sr from the two source mixing model including sea salt and bedrock weathering...............99 4.1 Flowchart summarizing the input data, parameterization steps and parameter estimation methods for the igneous submodel.....................................................................157 4.2 Terrane ages and boundaries across Alaska.................................................................. 158 4.3 Predicted 87Sr/86Sr in bedrock across Alaska................................................................159 4.4 Bedrock model validation................................................................................................ 160 4.5 Calibration of the Sr chemical weathering model.........................................................161 4.6 Map of sampling sites (with ID numbers) and their associated observed 87Sr/86Sr (color scale) for the catchment water model validation dataset modified from Brennan et al. (in press)..............................................................................................................................162 4.7 Catchment model validation for Alaska Rivers............................................................ 163 4.8 Predicted 87Sr/86Sr from the catchment water model applied across Alaska.............164 xi ACKNOWLEDGMENTS I am greatly endebtted to my advisor, Gabriel Bowen, who, through his example, his constructive reviews, his advice and his support has helped me to accomplish this work. He has been an incredibly devoted advisor and has helped me to enjoy my research and to more broadly love science. It could have been easy for such a brilliant young professor to give me little independence and to force his research interests on me, but instead, Gabe helped me to develop my own thinking in a friendly but efficient advising style. Gabe is a great example of what a good advisor should be for doctoral students. He has enough patience to let me think on my own and he fully accepted the mistakes and detours that came from leaving me so much room. I believe he gave me a solid background to face the challenges of the professional academic world. At a personal level, Gabe also gives an invaluable example of a balanced and happy man. He manages to be an amazing and hard-working scientist, a sociable man caring for others and a good father. His example and personal accomplishement were at least as valuable in my eyes as his professional career and I greatly admire the person he is. Finally, he and his wife, Brenda, were always supportive of my personal choices and of my family and contributed to make my PhD peacefull, happy and balanced. I thank my previous committee members from Purdue University (Matt Huber, Timothy Filley and Chi-Hua Huang) and my current committee members from the University of Utah: Lisa Stright, Thure Cerling, Brett Tipple and Diego Fernandez, who help me to think critically about this work and were always available for talking about science. I particularly thank Thure for his fascinating point of view on many science topics. I particularly thank Lisa for her inputs in geostatistics, her incredible teaching skills and her great example of raising young kids while being an accomplished scientist. I particularly thank Brett for teaching me gas chromatography methods and for fun discussions in the lab. I particuclarly thank Diego for initiating me into ICP-MS methods. I thank all the collaborators who made this work possible, particularly Jens Hartmann, Casey Kennedy, Sean Brennan, Jason Laffoon, Nils Moosdorf, Matthew Wooller and Justin Vandevelde as well as all the institutions who manage and made available the various data I have used throughout this PhD (Earthchem, GEOROC, USGS). I also thank all the staff, and lab technicians at Purdue and University and at the University of Utah for helping me in my work. I thank all the members of the IREH and SPATIAL laboratories for their friendship and help throughout this PhD, particularly Justin Vandevelde, Bianca Maibauer, Alex Lowe and Stephen Ruegg. I want to thank my parents in France for giving me such a wonderful education and encouraging my scientific curiosity, my brother and sisters for supporting me through those years away from my homeland and my family-in-law for adopting me and always making me feel at home in America. I mostly want to thank my wife, Hilary, who has been incredibly patient and helpful throughout those years of hardwork. Finally I want to thank my kids for making my PhD much more fun and challenging and for keeping my feet on the ground. Funding sources for this work are acknowleged at the end of each chapter. xiii CHAPTER I INTRODUCTION Isotope geochemistry in provenance studies In a globalized world, where movements of materials and individuals are accelerating and where consumers are increasingly concerned about product quality and provenance, tracing the origin of natural materials and human products is of great relevance. A growing number of international and/or local legislative initiatives in a variety of domains have helped to improve the traceability of human and natural products (e.g., food, endangered species) by introducing programs and labels to track provenance (e.g., United States "Country of Origin Labeling for Food Products"). The proliferation of those labels in recent years has increased the demand for new techniques able to resolve provenance at different spatial scales. Stable isotopes have gained considerable interest as a tool for provenance studies and have been successfully applied in a variety of fields to determine product authenticity for regulating trade practices (Kelly et al., 2005), to track the travel histories of individuals in criminal and/or archeological cases (Bentley, 2006), and to follow migration pathways of endangered and exploited animal populations (Hobson et al., 2010). The application of stable isotope geochemistry for resolving provenance relies on comparing the isotopic signature of a sample of unknown origin to that of baseline maps characterizing the isotopic signature of the potential isotope source(s) of the sample. Consequently, the application of stable isotopes as a tool of provenance requires mature and cost-effective analytical capabilities to analyze stable isotope ratio in the sample of interest, as well as mechanistic models able to predict the isotopic signatures of elemental sources at varied spatiotemporal scales. Oxygen (O) and hydrogen (H) isotope ratios are the most commonly applied isotope systems to resolve provenance in terrestrial environments (Bowen, 2010b). The 2 main advantages of applying O and H isotope ratios for provenance studies is that H and O stable isotope ratio analyses are precise, rapid and cost-effective and that the processes controlling the spatial isotopic variations of H and O on the Earth's surface are well-understood and can be used to develop accurate predictive models (Bowen, 2010b). The spatial variability of O and H isotope ratios in reservoirs of the Earth surface originates primarily from isotopic fractionation in biogeochemical processes as water cycles through different reservoirs (Bowen, 2010b). Isotopic fractionation imparts a unique spatiotemporal "isotopic label" or fingerprint to a given reservoir and/or material, and this unique fingerprint coupled with predictive models of isotope variations can be used to model the probability of geographic origin (Wunder, 2010). Decades of work have led to the development of models to predict the O and H isotope ratio variations in ocean water (LeGrande and Schmidt, 2006), rainfall (Bowen, 2010a), surface water (Bowen et al., 2011), tap water (Bowen et al., 2007) or animal tissues (West et al., 2007). Maps derived from those predictive models are widely distributed (Bowen et al., 2014) and routinely integrated to interpret provenance in fields as varied as forensics, archeology, atmospheric sciences, ecology, and paleoclimate. However, one fundamental limitation in applying O and H isotope in provenance studies is the generally broad scale of O and H isotope ratio variations. H and O isotope variations are primarily controlled by large scale (10 to 100 km) atmospheric processes that produce a continuous gradient in isotope variations and make geographic assignment nonunique (Bowen, 2010a; Farmer et al., 2008). 3 Strontium isotope geochemistry for provenance studies Strontium isotope ratio (87Sr/86Sr) constitutes an alternative and complementary tool to enhance the spatial resolution for provenance studies because strontium (Sr) in soils, waters, plants, and animals is sourced primarily from local bedrock. As such, 87Sr/86Sr patterns follow discrete variations of geological regimes with relatively constant 87Sr/86Sr within geological units of known age and lithology (Capo et al., 1998). This discrete patterning is superimposed by 87Sr/86Sr variability associated with local and regional geological processes. Local 87Sr/86Sr heterogeneity originates from variations in petrology, sedimentary provenance, bulk composition, or postburial alteration processes, whereas larger-scale 87Sr/86Sr heterogeneity are a function of regional tectonic or sedimentary basin processes. This multiscale and discrete patterning of 87Sr/86Sr variability is drastically different from the continuous patterns of O and H isotopes, providing complementary information to those isotopes for resolving provenance at multiple scales. Another critical advantage of the Sr isotope systems in comparison with traditional stable isotopes used in geoprofiling studies is that interpretation of 87Sr/86Sr variations are not complicated by isotopic fractionation (Capo et al., 1998). Small mass-dependent isotopic fractionation of Sr isotopes can occur in geologic and biological processes, but this isotopic fractionation is corrected for during mass spectrometric measurement by normalization of the nonradiogenic isotopes to known values (Capo et al., 1998). As a result, the measured 87Sr/86Sr reflects only variations in the amount of radiogenic 87Sr present in the sample, which ultimately is a function of its Sr source. Sr has also a long residence time in most reservoirs, which, combined with the absence of isotopic fractionation, leads to relatively constant 87Sr/86Sr signatures in bedrock and soils 4 at human timescales. These properties make 87Sr/86Sr a conservative tracer in Earth surface reservoirs, thus greatly simplifying interpretation of 87Sr/86Sr data. 5 Sr isotope systematic s Sr (atomic number 38) is a divalent alkaline Earth trace element which can substitute for Ca in Ca-bearing minerals such as plagioclase feldspar, apatite, sulfates (gypsum and anhydrite), and carbonates. Sr is one of the most abundant trace elements and is ubiquitous on the Earth's surface, making isotopic analysis relatively easy in comparison with other radiogenic isotopes. Sr has four naturally occurring stable isotopes 84Sr (0.56%), 86Sr (9.87%), 87Sr (7.04%) and 88Sr (82.53%) with 87Sr originating from the P-decay of rubidium 87 (87Rb) (decay constant ^=1.42*10-11 year-1) (Faure, 1977). 87Sr (daughter of 87Rb) is not concentrated in the same rock types as the other stable Sr isotopes because rubidium (Rb; atomic number 37) is an alkali metal and substitutes for K in K-bearing minerals such as muscovite, biotite, alkali feldspars (orthoclase and microcline), clays (illite) and evaporites (sylvite, carnallite). As a result, the present day quantity of 87Sr normalized to the naturally occurring and invariant 86Sr (87Sr/86Sr) in a given rock can be expressed using the radiogenic production equation as: ^ 7S r ^ 86 Sr V J r o c k 87Sr V 86sr j , + ^ 87 Rb^ 86 Sr V J r o c k (eM - 1) (11) where 87Sr/86Sr variations in rocks ( f 7S r / 6Sr)rock) are a function of: 1) the initial 87Sr/86Sr ( f 7Sr/i6Sr)l) which depends on the geological history of the parent rock, 2) bedrock age (t) which controls the fraction of 87Rb that decayed into 87Sr, and 3) the 87Rb/86Sr of the rock ((87R b /6Sr)rock) which varies with lithology because of the specific affinity of Rb and Sr with different minerals. This radiogenic equation is the basis for the rubidium-strontium dating method, used to determine the time of crystallization of igneous rocks such as granites (Faure, 1977). 87Rb/86Sr is directly proportional to the Rb/Sr and can be expressed as: 87Rb Rb(w87Rb)(mSr) ^ Rb , ( 1.2) 86Sr Sr(w*6Sr)(mRb) ~ ' Sr where m to refers to the atomic mass of an element and w to the abundance (%) of an isotope. Strontium cycle Sr is a relatively lithophilic element and has become increasingly concentrated on the more surficial layers of the Earth (Mantle and Crust) as geochemical differentiation processes (e.g., fractional crystallization) occurred during Earth's history. Sr is particularly concentrated in Ca-rich rock types such as carbonates, evaporites or intermediate igneous rock and is relatively less concentrated in felsic igneous rocks and siliciclastic sediments. Bedrock is the principal source of Sr to the Earth surface and Sr is exported from rocks to other reservoirs through erosion, weathering, and biological uptake (Capo et al., 1998). Erosion and weathering of bedrock and soils transfer Sr to the hydrosphere and the atmosphere in dissolved form or bounded to other particules (Capo et al., 1998). During the cycling of Sr on the Earth surface, part of this Sr is uptaken by the biosphere and is concentrated in certain Ca-rich biological tissues (e.g., bones). 6 7 Ultimately, Sr is transported to the ocean where it is deposited in carbonates on the ocean floor and recycled back into the Earth's interior through subduction (Capo et al., 1998). 87Sr/86Sr variations in the Earth throughout geological time All rock reservoirs inherited an identical initial 87Sr/86Sr signature (0.699) from the well-mixed primordial Earth (Wetherill et al., 1973). 87Sr/86Sr evolution of distinct geological reservoirs was induced by geochemical differentiation associated with fractional crystallization and the formation of different Earth layers and rock reservoirs (Faure, 1977). As geochemical differentiation progressed, Sr and even more so, Rb, concentrated into melts, resulting in high Rb/Sr ratios in the continental crust and its progressive diminution in the residual mantle (Faure, 1977). The 87Sr/86Sr of the bulk Earth increased as 87Rb decayed into 87Sr and this increase was enhanced in the crust as 87Rb became concentrated by recycling. As new rock reservoirs were formed from crustal or mantle precursors, they inherited the 87Sr/86Sr of their parent, but in most cases fractional crystallization led to dissimilar Rb/Sr, causing their 87Sr/86Sr to evolve along a different Rb/Sr slope than the parent material. Rocks with higher (lower) Rb/Sr than their parent evolved along a steeper (flatter) slope. The 87Sr/86Sr signatures of the parent and new rock reservoirs further diverged as time progressed because of the different 87Rb content of each reservoir. At equal Rb/Sr ratio, older rocks have higher 87 Sr/ 86Sr than younger rocks because 87Rb had more time to decay in the older reservoir. At equal age, rock with higher Rb/Sr have higher 87 Sr/ 86 Sr than rock with low Rb/Sr because more 87Rb 87 is available to decay into Sr. Combined effects of age and lithology explain the current first-order patterns of 87 Sr/ 86 Sr on the Earth surface, with high 87 Sr/ 86Sr in regions dominated with old felsic rocks (e.g., cratonic shields) and low 87Sr/86Sr in newly formed mafic terranes (e.g., Alaska volcanic terranes). Application of 87Sr/86Sr in low temperature geochemistry For decades, scientists have taken advantage of the well-documented 87Sr/86Sr variations in seawater to constrain the age of marine sediments (McArthur et al., 2001). The method relies on the assumption that 87Sr/86Sr of seawater depends primarily on the 87Sr/86Sr of the crust being altered at a given time period. 87Sr/86Sr analyses of marine carbonates from different time periods have helped to reconstruct a fairly complete, high resolution record of 87Sr/86Sr variation in seawater throughout Earth's history (Halverson et al., 2007; Shields and Veizer, 2002; Veizer, 1989). The 87Sr/86Sr signature in seawater displays a progressive increase with time associated with the increased felsic nature of the recycled crust. This background increase is overprinted by multimillion year time scale variations coinciding with different tectonic events, climate modes and supercontinent cycle stages (Veizer et al., 1999). The result is that the 87Sr/86Sr value and 87 86 the pattern of Sr/ Sr variations of a given time period are relatively diagnostic (little redundancy in 87Sr/86Sr values through time). Moreover, all carbonates deposited at a given time period have a similar 87Sr/86Sr signature (Veizer, 1989) because of the long residence time of Sr in seawater (Vollstaedt et al., 2014). Combined, those characteristics can be used for chronostratigraphy by correlating the 87Sr/86Sr of marine sediments of unknown age with the well-documented and time-constrained 87Sr/86Sr marine curve (McArthur et al., 2001). Another common application of 87 Sr/ 86 Sr is based on interpreting 87 Sr/8 6 Sr 8 variations in seawater as a tool to reconstruct the tectonic history of the Earth's surface (Veizer et al., 1999). Volumes of literature have been published to estimate the current Sr isotope budget and interpret 87Sr/86Sr shifts and trends in seawater throughout Earth history. The modern Sr isotope budget in the ocean depends primarily on changes in the magnitude and 87Sr/86Sr signature of a continental radiogenic flux of Sr from runoff and groundwater and a nonradiogenic ocean Sr fluxes from crust-seawater interactions (Veizer et al., 1999). The most recent estimates of those fluxes indicate that the radiogenic continental flux represents 59% of the total Sr flux - 43% of which originates from siliciclastic sediments (0.721) and 57% from carbonates (0.708) - whereas the nonradiogenic mafic flux represents 41% of the Sr flux - 27% of which are from oceanic crust seawater interactions (0.703) and 73% from volcanic arcs (0.7035) (Allegre et al., 2010). Based on those estimates, periods when 87Sr/86Sr in seawater are high are thought to represent time when continental weathering rates or 87Sr/86Sr signatures of the continental rocks were high. High continental weathering rates coincide with periods of orogenesis (Peters and Gaines, 2012; Raymo et al., 1988), periods of equatorial positioning of plate tectonics (Godderis et al., 2014) or periods of climate shifts from greenhouse to icehouse (Zachos et al., 1999). High 87Sr/86Sr signatures of the continental surface occur during periods of intense crustal recycling during collisional orogenesis 87 86 (Condie and Aster, 2013). Periods when Sr/ Sr in seawater is low are thought to reflect periods of formation of nonradiogenic continental crust (Condie and Aster, 2013), increase in seafloor spreading rates (Graham et al., 1982) or occurrence of massive continental floodbasalt and continental arcs (Allegre et al., 2010; Das and Krishnaswami, 2007). 9 In the last decades, applications of 87Sr/86Sr for provenance of organic and nonorganic materials have been rapidly extending, at least in part due to the development of high performance laser ablation multicollector inductively coupled plasma mass spectrometry, which allows rapid and high precision analysis and requires very small amounts of sample. Those technological advances have opened the door to new applications for Sr isotope geochemistry, including: 1) reconstructing the migration pathways of mammals (Koch et al., 1995), birds (Sellick et al., 2009), fresh water, anadromous fishes (Barnett-Johnson et al., 2008; Kennedy et al., 2005; Walther et al., 2011) and paleofauna (Britton et al., 2011; Hoppe et al., 1999; Price et al., 2002), 2) determining feeding habits of ancient humans (Copeland et al., 2011) and animals (Feranec et al., 2007; Radloff et al., 2010), 3) distinguishing region of origin of agricultural or natural products such as rice (Kawasaki et al., 2002), wine (Marchionni et al., 2013), water (Voerkelius et al., 2010), milk (Crittenden et al., 2007) or illegal drugs (West et al., 2009), 4) identifying nonlocal individuals in archeology and forensic cases (Bentley, 2006; Price et al., 2002; Schroeder et al., 2009), 5) determining the geographic source of dust aerosols (Grousset and Biscaye, 2005; Kurtz et al., 2001), 6) reconstructing silicate and carbonate weathering patterns in watersheds (Blum et al., 1998; Gaillardet et al., 1997; Horton et al., 1999; Huh and Edmond, 1999; Huh et al., 1998a; Huh et al., 1998b; Millot et al., 2002; Millot et al., 2003; Pretti and Stewart, 2002; Probst et al., 2000; Rad et al., 2007), 7) identifying seasonal variations in sources of elements to river (Douglas et al., 2013; Nakano and Tanaka, 1997; Voss et al., 2014) and, 8) distinguishing element sources to soils and ecosystems (Bern et al., 2005; Chadwick et al., 2009; Pett-Ridge et al., 2009). In most of the current applications of Sr isotope 10 geochemistry for determining provenance, the spatial 87Sr/86Sr variations of the Sr sources are poorly constrained, particularly when the potential Sr sources cover large spatial scales (e.g., water or food sources for animals with large foraging areas). The application and interpretation of Sr isotope geochemistry in provenance studies at regional spatial scales requires the development of models predicting the 87Sr/86Sr signature of Sr sources. Modeling 87Sr/86Sr variations Several approaches have been tested in recent years to develop regional scale predictive 87Sr/86Sr maps of potential Sr sources including bedrock (Beard and Johnson, 2000), river water (Hegg et al., 2013) or bioavailable Sr (Frei and Frei, 2011; Price et al., 2002). The most common approach used in ecology and archeology is to use interpolation to generate a map of 87Sr/86Sr based on analyses of a reference substrate that record or approximate the 87Sr/86Sr of the Sr sources. For instance, in archeology, interpolated maps representing the 87Sr/86Sr variations of the "bioavailable Sr reservoir" have been derived from 87Sr/86Sr analyses of local surface water (Frei and Frei, 2011), local flora (Price et al., 2002) or local fauna (Bentley and Knipper, 2005; Hodell et al., 2004; Laffoon et al., 2012). While this method can give precise 87Sr/86Sr prediction at local scale, it is hampered by the challenge of selecting appropriate reference substrates, which is nontrivial as different sample materials may integrate different spatial and temporal scales of 87Sr/86Sr variation. Ultimately, applying this method at large spatial scale is also data intensive and costly, and in most cases traditional interpolation algorithms are unable to explicitly consider the discrete patterning of 87Sr/86Sr variations. 11 Another approach applied to derive predictive 87Sr/86Sr models consists of training a multiple linear regression model (MLRM) using lithology and age as predictor. This method has the advantage of partly accounting for processes governing the spatial structure of 87Sr/86Sr variations (Hegg et al., 2013). However, the dependence of 87Sr/86Sr on a large number of nonindependent predictors (age and multiple lithological classes) limits the performance of MLRM and requires a large number of 87Sr/86Sr analyses to make the model statistically robust. The MLRM approach appears to function relatively well in areas where lithological and geological complexity is low, limiting the number of potential predictors. However, the method becomes increasingly uncertain for areas with more complex geology where both age and lithology control 87Sr/86Sr variations. A more promising and complementary approach to developing a unified framework to predict 87Sr/86Sr at large spatial scales is to develop process-based spatial models. This approach was initiated by Beard and Johnson (2000) who developed a model predicting 87Sr/86Sr in bedrock over the conterminous USA. Their method is based on the assumption that bedrock age is the primary control of 87Sr/86Sr variations, because rock age determines the fraction of 87Rb that has decayed into 87Sr. The predictive power of this model was relatively low, explaining roughly 30% of the total variance for both bedrock and water (Chesson et al., 2012; Hobson et al., 2010). Despite the poor predictive power of this initial model, Beard and Johnson's (2000) idea of using geological maps to predict 87Sr/86Sr was the inspiration for our work and laid the foundation for the development of new, more accurate process-based 87Sr/86Sr models. 12 Objectives and outline In this dissertation, we present a series of modeling activities aimed at establishing a basis for process-based predictive GIS modeling of Sr isotopes distribution at continental to global scale. The approach leverages publically available geospatial data on rock geochemistry, surficial and bedrock geology, climate, hydrology, and aerosols to model the input and propagation of Sr from multiple geological sources through hydrosystems and ecosystems. In the second chapter, we formulate and calibrate a bedrock model which predicts 87Sr/86Sr in bedrock as a function of rock age and lithology. This initial bedrock model is coupled to a basic chemical weathering model to predict 87Sr/86Sr in soil and river waters. This initial chemical weathering model calculates the export of Sr from rock to river water as a function of the difference in weathering rates and Sr content of different rock types. The performance of the model is tested against rock, water and biological data over the conterminous USA and demonstrates encouraging performance. In the third chapter, we build upon our initial modeling effort to formulate a bioavailable Sr model which explicitly predicts the 87Sr/86Sr in biological material and accounts for the mixing of Sr from multiple sources in ecosystems. This model is used to predict the 87Sr/86Sr of bioavailable Sr for the circum- Caribbean region, and significantly improves the predictive power of our models when tested against biological datasets. In the final chapter, we revisit the bedrock and water models by developing a unified flexible geostatistical framework to predict 87Sr/86Sr in Alaska rivers. In this final step, we improve the calibration of the bedrock model to predict local scale 87Sr/86Sr variability and represent prediction uncertainty. This new bedrock model is coupled to a 13 14 fully independent, Sr-specific chemical weathering model predicting the spatial variations in the rate of Sr release from rocks. The model is used to predict Sr and 87Sr/86Sr in rivers 87 86 of Alaska and is tested against an independent dataset of Sr/ Sr analyses from streams and rivers of Alaska. References Allegre, C.J. et al., 2010. The fundamental role of island arc weathering in the oceanic Sr isotope budget. Earth and Planetary Science Letters, 292(1-2): 51-56. Balasse, M., Ambrose, S.H., Smith, A.B., Price, T.D., 2002. The seasonal mobility model for prehistoric herders in the South-western Cape of South Africa assessed by isotopic analysis of sheep tooth enamel. Journal of Archaeological Science, 29(9): 917-932. Barnett-Johnson, R., Pearson, T.E., Ramos, F.C., 2008. Tracking natal origins of salmon using isotopes, otoliths, and landscape geology. Limnology and Oceanography, 53(4): 1633-1642. Beard, B.L., Johnson, C.M., 2000. Strontium isotope composition of skeletal material can determine the birth place and geographic mobility of humans and animals. Journal of Forensic Sciences, 45(5): 1049-1061. Bentley, R.A., 2006. Strontium isotopes from the earth to the archaeological skeleton: A review. Journal of Archaeological Method and Theory, 13(3): 135-187. Bentley, R.A., Knipper, C., 2005. Geographical patterns in biologically available strontium, carbon and oxygen isotope signatures in prehistoric SW Germany. Archaeometry, 47: 629-644. Bern, C.R., Townsend, A.R., Farmer, G.L., 2005. Unexpected dominance of parent-material strontium in a tropical forest on highly weathered soils. Ecology, 86(3): 626-632. Blum, J.D., Gazis, C.A., Jacobson, A.D., Page Chamberlain, C., 1998. Carbonate versus silicate weathering in the Raikhot watershed within the High Himalayan Crystalline Series. Geology, 26(5): 411-414. Bowen, G.J., 2010a. Statistical and Geostatistical Mapping of Precipitation Water Isotope Ratios. In: West, J.B., Bowen, G.J., Dawson, T.E., Tu, K.P. (Eds), Isoscapes. Springer Netherlands, pp. 139-160. 15 Bowen, G.J., 2010b. Isoscapes: Spatial pattern in isotopic biogeochemistry. Annual Review of Earth and Planetary Sciences, 38(1): 161-187. Bowen, G.J., Ehleringer, J.R., Chesson, L.A., Stange, E., Cerling, T.E., 2007. Stable isotope ratios of tap water in the contiguous United States. Water Resources Research, 43(3): W03419. Bowen, G.J., Kennedy, C.D., Liu, Z., Stalker, J., 2011. Water balance model for mean annual hydrogen and oxygen isotope distributions in surface waters of the contiguous United States. Journal of Geophysical Research: Biogeosciences, 116(G4): G04011. Bowen, G.J., Liu, Z., Vander Zanden, H.B., Zhao, L., Takahashi, G., 2014. Geographic assignment with stable isotopes in IsoMAP. Methods in Ecology and Evolution, 5(3): 201-206. Britton, K. et al., 2011. Strontium isotope evidence for migration in late Pleistocene rangifer: Implications for Neanderthal hunting strategies at the middle Palaeolithic site of Jonzac, France. Journal of Human Evolution, 61(2): 176-185. Capo, R.C., Stewart, B.W., Chadwick, O.A., 1998. Strontium isotopes as tracers of ecosystem processes: Theory and methods. Geoderma, 82(1-3): 197-225. Chadwick, O.A., Derry, L.A., Bern, C.R., Vitousek, P.M., 2009. Changing sources of strontium to soils and ecosystems across the Hawaiian Islands. Chemical Geology, 267(1-2): 64-76. Chesson, L.A. et al., 2012. Strontium isotopes in tap water from the coterminous USA. Ecosphere 3:art67. http://dx.doi.org/10.1890/ES12-00122.! Condie, K.C., Aster, R.C., 2013. Refinement of the supercontinent cycle with Hf, Nd and Sr isotopes. Geoscience Frontiers, 4(6): 667-680. Copeland, S.R. et al., 2011. Strontium isotope evidence for landscape use by early hominins. Nature, 474(7349): 76-78. Crittenden, R.G. et al., 2007. Determining the geographic origin of milk in Australasia using multi-element stable isotope ratio analysis. International Dairy Journal, 17(5): 421-428. Das, A., Krishnaswami, S., 2007. Elemental geochemistry of river sediments from the Deccan Traps, India: Implications to sources of elements and their mobility during basalt-water interaction. Chemical Geology, 242(1-2): 232-254. Douglas, T.A., Blum, J.D., Guo, L., Keller, K., Gleason, J.D., 2013. Hydrogeochemistry of seasonal flow regimes in the Chena River, a subarctic watershed draining 16 discontinuous permafrost in interior Alaska (USA). Chemical Geology, 335(0): 48-62. Farmer, A., Cade, B.S., Torres-Dowdall, J., 2008. Fundamental limits to the accuracy of deuterium isotopes for identifying the spatial origin of migratory animals. Oecologia, 158(2): 183-192. Faure, G., 1977. Principles of Isotope Geology. Second edition, John Wiley and Sons, New York, 589 p. Feranec, R., Hadly, E., Paytan, A., 2007. Determining landscape use of Holocene mammals using strontium isotopes. Oecologia, 153(4): 943-950. Frei, K.M., Frei, R., 2011. The geographic distribution of strontium isotopes in Danish surface waters - A base for provenance studies in archaeology, hydrology and agriculture. Applied Geochemistry, 26(3): 326-340. Gaillardet, J., Dupre, B., Allegre, C.J., Negrel, P., 1997. Chemical and physical denudation in the Amazon River Basin. Chemical Geology, 142(3-4): 141-173. Godderis, Y., Donnadieu, Y., Le Hir, G., Lefebvre, V., Nardin, E., 2014. The role of palaeogeography in the Phanerozoic history of atmospheric CO2 and climate. Earth-Science Reviews, 128(0): 122-138. Graham, D.W., Bender, M.L., Williams, D.F., Keigwin Jr, L.D., 1982. Strontium-calcium ratios in Cenozoic planktonic foraminifera. Geochimica et Cosmochimica Acta, 46(7): 1281-1292. Grousset, F.E., Biscaye, P.E., 2005. Tracing dust sources and transport patterns using Sr, Nd and Pb isotopes. Chemical Geology, 222(3-4): 149-167. Halverson, G.P., Dudas, F.O., Maloof, A.C., Bowring, S.A., 2007. Evolution of the 87Sr/86Sr composition of Neoproterozoic seawater. Palaeogeography, Palaeoclimatology, Palaeoecology, 256(3-4): 103-129. Hegg, J.C., Kennedy, B.P., Fremier, A.K., 2013. Predicting strontium isotope variation and fish location with bedrock geology: Understanding the effects of geologic heterogeneity. Chemical Geology, 360-361(0): 89-98. Hobson, K., Barnett-Johnson, R., Cerling, T., 2010. Using Isoscapes to Track Animal Migration. In: West, J.B., Bowen, G.J., Dawson, T.E., Tu, K.P. (Eds.), Isoscapes. Springer Netherlands, pp. 273-298. Hodell, D.A., Quinn, R.L., Brenner, M., Kamenov, G., 2004. Spatial variation of strontium isotopes (Sr-87/Sr-86) in the Maya region: A tool for tracking ancient human migration. Journal of Archaeological Science, 31(5): 585-601. 17 Hoppe, K.A., Koch, P.L., Carlson, R.W., Webb, S.D., 1999. Tracking mammoths and mastodons: Reconstruction of migratory behavior using strontium isotope ratios. Geology, 27(5): 439-442. Horton, T.W., Chamberlain, C.P., Fantle, M., Blum, J.D., 1999. Chemical weathering and lithologic controls of water chemistry in a high-elevation river system: Clark's Fork of the Yellowstone River, Wyoming and Montana. Water Resources Research, 35(5): 1643-1655. Huh, Y., Edmond, J.M., 1999. The fluvial geochemistry of the rivers of Eastern Siberia: III. Tributaries of the Lena and Anabar draining the basement terrain of the Siberian Craton and the Trans-Baikal Highlands. Geochimica et Cosmochimica Acta, 63(7-8): 967-987. Huh, Y., Panteleyev, G., Babich, D., Zaitsev, A., Edmond, J.M., 1998a. The fluvial geochemistry of the rivers of Eastern Siberia: II. Tributaries of the Lena, Omoloy, Yana, Indigirka, Kolyma, and Anadyr draining the collisional/accretionary zone of the Verkhoyansk and Cherskiy ranges. Geochimica et Cosmochimica Acta, 62(12): 2053-2075. Huh, Y., Tsoi, M.Y., Zaitsev, A., Edmond, J.M., 1998b. The fluvial geochemistry of the rivers of Eastern Siberia: I. Tributaries of the Lena River draining the sedimentary platform of the Siberian Craton. Geochimica et Cosmochimica Acta, 62(10): 1657-1676. Kawasaki, A., Oda, H., Hirata, T., 2002. Determination of strontium isotope ratio of brown rice for estimating its provenance. Soil Science and Plant Nutrition, 48(5): 635-640. Kelly, S., Heaton, K., Hoogewerff, J., 2005. Tracing the geographical origin of food: The application of multi-element and multi-isotope analysis. Trends in Food Science & Technology, 16(12): 555-567. Kennedy, B.P., Chamberlain, C.P., Blum, J.D., Nislow, K.H., Folt, C.L., 2005. Comparing naturally occurring stable isotopes of nitrogen, carbon, and strontium as markers for the rearing locations of Atlantic salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences, 62(1): 48-57. Koch, P.L., Halliday, A.N., Water, L.M., Stearley, R.F., Huston, T.J., Smith, G.R., 1992. Sr isotopic composition of hydroxyapatite from recent and fossil salmon: The record of lifetime migration and diagenesis. Earth and Planetary Science Letters, 108(4): 277-287. Koch, P.L., Heisinger, J., Moss, C., Carlson, R.W., Fogel, M.L., Behrensmeyer, A.K., 1995. Isotopic tracking of change in diet and habitat use in African elephants. Science, 267(5202): 1340-1343. 18 Kurtz, A.C., Derry, L.A., Chadwick, O.A., 2001. Accretion of Asian dust to Hawaiian soils: Isotopic, elemental, and mineral mass balances. Geochimica Et Cosmochimica Acta, 65(12): 1971-1983. Laffoon, J.E., Davies, G.R., Hoogland, M.L.P., Hofman, C.L., 2012. Spatial variation of biologically available strontium isotopes (87Sr/86Sr) in an archipelagic setting: A case study from the Caribbean. Journal of Archaeological Science, 39(7): 2371­2384. LeGrande, A.N., Schmidt, G.A., 2006. Global gridded data set of the oxygen isotopic composition in seawater. Geophysical Research Letters, 33(12): L12604. Marchionni, S., Braschi, E., Tommasini, S., Bollati, A., Cifelli, F., Mulinacci, N., Mattei, M., Conticelli, S., 2013. High-precision 87Sr/86Sr analyses in wines and their use as a geological fingerprint for tracing geographic provenance. Journal of Agricultural and Food Chemistry, 61(28): 6822-6831. McArthur, J.M., Howarth, R.J., Bailey, T.R., 2001. Strontium isotope stratigraphy: LOWESS Version 3: Best fit to the marine Sr-isotope curve for 0-509 Ma and accompanying look-up table for deriving numerical age. The Journal of Geology, 109(2): 155-170. Millot, R., Gaillardet, J., Dupre, B., Allegre, C.J., 2002. The global control of silicate weathering rates and the coupling with physical erosion: New insights from rivers of the Canadian Shield. Earth and Planetary Science Letters, 196(1-2): 83-98. Millot, R., Gaillardet, J.e., Dupre, B., Allegre, C.J., 2003. Northern latitude chemical weathering rates: Clues from the Mackenzie River Basin, Canada. Geochimica et Cosmochimica Acta, 67(7): 1305-1329. Nakano, T., Tanaka, T.-i., 1997. Strontium isotope constraints on the seasonal variation of the provenance of base cations in rain water at Kawakami, Central Japan. Atmospheric Environment, 31(24): 4237-4245. Peters, S.E., Gaines, R.R., 2012. Formation of the /'Great Unconformity/' as a trigger for the Cambrian explosion. Nature, 484(7394): 363-366. Pett-Ridge, J.C., Derry, L.A., Kurtz, A.C., 2009. Sr isotopes as a tracer of weathering processes and dust inputs in a tropical granitoid watershed, Luquillo Mountains, Puerto Rico. Geochimica Et Cosmochimica Acta, 73(1): 25-43. Pretti, V.A., Stewart, B.W., 2002. Solute sources and chemical weathering in the Owens Lake watershed, eastern California. Water Resources Research, 38(8):1-18. Price, T.D., Burton, J.H., Bentley, R.A., 2002. The characterization of biologically available strontium isotope ratios for the study of Prehistoric migration. 19 archaeometry, 44(1): 117-135. Probst, A., El Gh'mari, A., Aubert, D., Fritz, B., McNutt, R., 2000. Strontium as a tracer of weathering processes in a silicate catchment polluted by acid atmospheric inputs, Strengbach, France. Chemical Geology, 170(1-4): 203-219. Rad, S.D., Allegre, C.J., Louvat, P., 2007. Hidden erosion on volcanic islands. Earth and Planetary Science Letters, 262(1-2): 109-124. Radloff, F.G.T., Mucina, L., Bond, W.J., Roux, P.J., 2010. Strontium isotope analyses of large herbivore habitat use in the Cape Fynbos region of South Africa. Oecologia, 164(2): 567-578. Raymo, M.E., Ruddiman, W.F., Froelich, P.N., 1988. Influence of late Cenozoic mountain building on ocean geochemical cycles. Geology, 16(7): 649-653. Saito, H., Goovaerts, P., 2000. Geostatistical interpolation of positively skewed and censored data in a dioxin-contaminated site. Environmental Science & Technology, 34(19): 4228-4235. Schroeder, H., O'Connell, T.C., Evans, J.A., Shuler, K.A., Hedges, R.E.M., 2009. Trans­Atlantic slavery: Isotopic evidence for forced migration to Barbados. American Journal of Physical Anthropology, 139(4): 547-557. Sellick, M.J., Kyser, T.K., Wunder, M.B., Chipley, D., Norris, D.R., 2009. Geographic variation of strontium and hydrogen isotopes in avian tissue: Implications for tracking migration and dispersal. Plos One, 4(3): e4735. Shields, G., Veizer, J., 2002. Precambrian marine carbonate isotope database: Version 1.1. Geochemistry Geophysics Geosystems, 3(6): 10.1029/2001GC000266. Veizer, J., 1989. Strontium isotopes in seawater through time. Annual Review of Earth and Planetary Sciences, 17(1): 141-167. Veizer, J. et al., 1999. Sr-87/Sr-86, delta C-13 and delta O-18 evolution of Phanerozoic seawater. Chemical Geology, 161(1-3): 59-88. Voerkelius, S. et al., 2010. Strontium isotopic signatures of natural mineral waters, the reference to a simple geological map and its potential for authentication of food. Food Chemistry, 118(4): 933-940. Vollstaedt, H. et al., 2014. The Phanerozoic S88/86Sr record of seawater: New constraints on past changes in oceanic carbonate fluxes. Geochimica et Cosmochimica Acta, 128(0): 249-265. Voss, B.M. et al., 2014. Tracing river chemistry in space and time: Dissolved inorganic 20 constituents of the Fraser River, Canada. Geochimica et Cosmochimica Acta, 124(0): 283-308. Walther, B.D., Dempster, T., Letnic, M., McCulloch, M.T., 2011. Movements of diadromous fish in large unregulated tropical rivers inferred from geochemical tracers. PloS One, 6(4): e18351. West, J.B., Ehleringer, J.R., Cerling, T.E., 2007. Geography and vintage predicted by a novel GIS model of wine S18O. Journal of Agricultural and Food Chemistry, 55(17): 7075-7083. West, J.B., Hurley, J.M., Dudas, F.O., Ehleringer, J.R., 2009. The stable isotope ratios of marijuana. II. Strontium isotopes relate to geographic origin. Journal of Forensic Sciences, 54(6): 1261-1269. Wetherill, G.W., Mark, R., Lee-Hu, C., 1973. Chondrites: Initial strontium-87/strontium- 86 ratios and the early history of the solar system. Science, 182(4109): 281-283. Wunder, M., 2010. Using Isoscapes to Model Probability Surfaces for Determining Geographic Origins. In: West, J.B., Bowen, G.J., Dawson, T.E., Tu, K.P. (Eds.), Isoscapes. Springer Netherlands, pp. 251-270. Zachos, J.C., Opdyke, B.N., Quinn, T.M., Jones, C.E., Halliday, A.N., 1999. Early cenozoic glaciation, antarctic weathering, and seawater 87Sr/86Sr: Is there a link? Chemical Geology, 161(1-3): 165-180. CHAPTER II MAPPING 87Sr/86Sr VARIATIONS IN BEDROCK AND WATER FOR LARGE SCALE PROVENANCE STUDIES Reprinted from Chemical Geology, Vol. 304-305, C.P., Bataille, G.J., Bowen, Mapping 87Sr/86Sr variations in bedrock and water for large scale provenance studies, pp. 39-52. Copyright 2012, with permission from Elsevier. 22 Chemical Geology 304-305 (2012) 39-52 ELSEVIER Contents lists available at SciVerse ScienceDirect Chemical Geology jo u rn a l h om e p a g e : w w w .e ls e v ie r .c o m / lo c a te /c h e m g e o Research papers 87c v /86 Mapping Sr/ Sr variations in bedrock and water for large scale provenance studies Clement P. Bataille *, Gabriel J. Bowen Department o f Earth and Atmospheric Sciences, Purdue University, West Lafayette, Indiana 4 7 9 0 7USA A R T I C L E I N F O Article history: Received 1 June 2011 Received in revised form 23 January 2012 Accepted 24 January 2012 Available online 31 January 2012 Editor: J.D. Blum Keywords: Strontium rubidium method Isoscape GIS Strontium isotope ratio Provenance A B S T R A C T Although variation in Sr/ Sr has been widely pursued as a tracer of provenance in environmental studies, forensics, archeology and food traceability, accurate methods for mapping variations in environmental 87Sr/ 86Sr at regional scale are not available. In this paper, we build upon earlier efforts to model 87Sr/86Sr in bed­rock by developing GIS-based models for Sr isotopes in rock and water that include the combined effects of lithology and time. Using published data, we fit lithology-specific model parameters for generalized equa­tions describing the concentration of radiogenic Sr in silicate and carbonate rocks. The new model explained more than 50% of the observed variance in measured Sr isotope values from independent global databases of igneous, metaigneous, and carbonate rocks, but performed more poorly (explaining 33% of the variance) for sedimentary and metasedimentary rocks. In comparison, a previously applied model formulation that did not include lithology-specific parameters explained only 20% and 8% of the observed variance for igneous and sedimentary rocks, respectively, and exhibited an inverse relationship with measured carbonate rock values. Building upon the bedrock model, we also developed and applied equations to predict the contribution of dif­ferent rock types to 87Sr/8SSr variations in water as a function of their weathering rates and strontium con­tent. The resulting water model was compared to data from 68 catchments and shown to give more accurate predictions of stream water 87Sr/86Sr (R2 = 0.70) than models that did not include lithological weathering parameters. We applied these models to produce maps ("isoscapes") predicting 87Sr/86Sr in bed­rock and water across the contiguous USA, and compared the mapped Sr isotope distributions to data on Sr isotope ratios of US marij'uana crops. Although the maps produced here are demonstrably imperfect and leave significant scope for further refinement, they provide an enhanced framework for lithology-based Sr isotope modeling and offer a baseline for provenance studies by constraining the 87Sr/86Sr in strontium sources at regional scales. © 2012 Elsevier B.V. All rights reserved. 1. In tro d u c tio n Strontium isotope ratio measurements (87Sr/86Sr) have been ap­plied in a wide variety of geoscience studies including chronostatigra-phy of marine sediments (Veizer et aL, 1999), petrology of igneous rocks (DePaolo, 1981), cation provenance and mobility (Chaudhuri and Clauer, 1993; Miller et al., 1993; Grousset and Biscaye, 2005; Chadwick et al., 2009), and quantitative models of chemical weather­ing (Clow et aL, 1997; Horton et al, 1999), More recently the use of 87Sr/86Sr has been extended to a wide range of new applications in hydrology (Hogan et al., 2000), forensics (Beard and Johnson, 2000; West et aL, 2009), archeology (Hodell et al., 2004; Bentley et al., 2008), ecology (Koch et al.t 1995; Chamberlain et al., 1997; Hoppe et al., 1999; Barnett-Johnson et al., 2008) and food traceability (Kelly et al., 2005; Crittenden et al., 2007; Voerkelius et al, 2010). These applications are based on the principle that 87Sr/86Sr of natural materials reflects the sources of strontium (Sr) available during their ! Corresponding author. Tel: + 1 765 404 4772; fax: E-mail address: [e-mail] (CP. Bataille). -1 765 4961210. formation (Dasch, 1969). For instance, in studies of animal prove­nance, the 87Sr/86Sr of the Sr assimilated in animal tissues reflects the different sources of ingested Sr obtained from water and/or food (Graustein, 1989). As a consequence, variations in 87Sr/86Sr of these tissues can be used to trace migration or changes in diet habits of a given organism (Capo et aL, 1998). Interpreting the 87Sr/86Sr signa­ture for provenance studies requires constraining the 87Sr/86Sr varia­tions of potential environmental sources of Sr. In this work, we attempt to model the spatial variations of 87Sr/86Sr in bedrock and water, two important sources of Sr to biological systems. The use of 87Sr/86Sr as a tracer is of particular interest because un­like for isotopes of the light elements, biological and instrumental mass-dependent fractionations are automatically corrected during measurements and thus, the 87Sr/86Sr directly reflects the Sr of the source. In addition their wide amplitude of variation on both large and small scales, low temporal variability, and relative high abun­dance for a trace element make Sr isotopes a strong candidate for tracing inorganic and organic materials, either independently or in conjunction with isotopic data from lighter elements (Graustein and Armstrong, 1983; Kawasaki et al., 2002; Bowen et al., 2005; Bowen, 2010). 0009-2541/$ - see front matter © 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.chemgeo.2012.01.02S 23 CP. Bataiiie, GJ. Bowen / Chemical Geology 304-305 (2012) 39-52 Several different approaches have been taken to map Sr iso topic variations at large scale. The fundamental theory underlying Sr iso­tope variation in geological material was summarized by Faure (1977), who proposed two equations describing the evolution of 87 Sr/&6Sr in mantle and crustal rocks that remain the basis for modeling 87Sr/86Sr in bedrock. Although several efforts were made to map 87Sr/86Sr available to ecosystems over local to regional scales based on field measurements in water, soils and organisms (Price et al, 1994; Ezzo et al., 1997; Hodell et al., 2004; Bentley and Knipper, 2005), Beard and Johnson (2000) made the first attempt to model 87Sr/86Sr variations in bedrock over large spatial scales. In their work, Beard and Johnson simplified Faure's theory by consider­ing rock age to be the only determinant of 87Sr/86Sr variations and mapped the 87Sr/86Sr in the USA based on rock unit ages reported in a dig­ital geological map. Their study suggested strongly patterned 87Sr/86Sr variations at continental scales but did not include systematic verification from field measurements. Nor did the authors advocate the use of their ‘fist-pass1 model for quantitative prediction of Sr isotope ratios. Because bedrock weathering is the ultimate source of Sr to biological systems, however, these authors proposed that with improved understanding the modeled patterns could be used to interpret the geographic origin of biological materials. Although bedrock Sr is the ultimate source of Sr to Earth surface systems, its isotopic composition can differ substantially from that of soils, surface water and organisms due to factors such as variation in weathering rates for different minerals or inputs from other sources such as atmospheric aerosols (Capo et al., 1998; Stewart et al., 1998; Bentley, 2006; Chadwick et al., 2009). For constraining the isotopic variations in source of Sr for provenance studies it is most appropriate to model the "biologically available Sr" as an ap­proximation of the Sr actually assimilated by organisms (Sillen et aL, 1998; Price et alM 2002; Hodell et al., 2004; Frei and Frei, 2011). In this regard, a theoretical steady state and time dependent model predicting 87Sr/86Sr evolution in "biologically available Sr" has been developed (Stewart et al., 1998). This model details the po­tential factors and sources causing transfers of Sr from soil to water and from water to ecosystems, but its applicability remains limited at regional scale because of the large number of variables to be con­strained. Furthermore, this model relies on empirical measurements to obtain the &7Sr/86Sr variations in bedrock (Stewart et al., 1998). An alternative approach to mapping spatial Sr isotope variation has been proposed by the TRACE project, which developed an empir­ical model for predicting 87Sr/86Sr in groundwater based on the mea­surement of 650 different European natural mineral waters (Voerkelius et al., 2010). Using this dataset and a geological map of Europe, a mean 87Sr/86Sr was calculated for each geological unit un­derlying sampled waters, and these values were extrapolated to sim­ilar rock units to develop a comprehensive prediction map for the continent. Although this model was shown to reproduce the large scale patterns of 87Sr/86Sr variations in biological materials, it re­quires iterative subjective analysis of regional geological and Sr iso­tope data and as a result is not immediately generalizable to other regions. Furthermore, the prediction accuracy of this approach is highly dependent on the density of sampling, and improving the res­olution or extending the spatial coverage will require expensive field campaigns. These previous efforts illustrate the difficulties of mapping 87Sr/ 86Sr variation at different scales in different sources. In this paper, we focus on developing scalable spatial 87Sr/86Sr predictions for bed­rock and water. We build upon the effort of Beard and Johnson to map 87Sr/86Sr in bedrock and we developed a simplified model of Sr cy­cling (Stewart et al., 1998) to extend 87Sr/86Sr prediction to water and ecosystems. We validate the models by using existing 87Sr/86Sr measurements. The resulting models provide baseline predictions for rock and water 87Sr/86Sr across the contiguous USA, and can be used over a range of spatial resolutions depending on the application of interest. Although these models are demonstrably imperfect and incomplete and their predictive power limited with respect to that desired in many potential applications, this work represents an im­portant step towards developing systematic spatial predictions for Sr isotopes with wide geochemical applications. 2. Bed ro ck m o d e ls Model derivation, calibration, and validation are described in the following sections. Additional details and documentation are avail­able in the accompanying Supplementary material. 2.1. Silicate model theory 87Sr production in rocks results from the radioactive decay of 87Rb, which decays to 87Sr with a half-life of 49 billion years. In a closed system, the ratio of radiogenic 87Sr to the stable isotope 86Sr in rocks slowly increases with time (t) as a function of the rock's Rb/Sr ratio: Sr mST = (i) where X is the decay constant of the parent isotope (1.42* 10" 11 yr_1) and (87Sr/86Sr)j is the initial 87Sr/86Sr (Faure, 1977). Rb/Sr varies between different layers and different rocks be­cause geochemical processes fractionate Rb and Sr due to the specific affinity of each element for different minerals (Carlson, 2003). Rb substitutes better for potassium (K) and Sr for calcium (Ca) in min­erals, and Rb/Sr tends to be high in felsic and K-bearing sedimentary rocks and low in mafic and carbonate rocks (Rudnick, 2003). The dis­similar affinities of Rb and Sr cause Rb/Sr to vary at large scales be­tween the mantle and the crust and at small scales between different rocks and minerals (Rudnick, 2003). Our model attempts to trace the evolution of 87Sr/86Sr values in silicate rocks from the time of Earth's formation to present (Fig. 1). We consider that the geological history of each rock started 4.5 billion years ago in a chemically homogeneous Earth. At that time, the Rb and Sr present in the earth were well-mixed and the 87Sr/86Sr is esti­mated to have been 0.699 based on measurements of chondrites (Wasserbu et al., 1969). Prior to crustal differentiation, the 87Sr/86Sr evolved slowly but homogeneously in the bulk earth. Following dif­ferentiation, different crustal layers inherited a higher Rb/Sr than the mantle due to the higher affinity of Rb for crustal minerals. Conse­quently, as the crust evolved, its 87Sr/86Sr deviated from the residual mantle value (Faure, 1977). As new rocks were formed from crustal 0.707 - J J s z ■ 13 / U J CRUSTS ■ BULK EARTH MANTLE H igh R b /S r A L ow R b/S r 3 2 Age (Ga) Fig. 1. Three-stage model for the evolution of S7Sr/86Sr in Earth materials through geological time. S7Sr accumulates in all pools due to 37Rb decay, with the rate depending of the Rb/Sr of each lithology. Gran it e l and Granite 2 are examples of rock formation occurring at different time during earth history. Modified from Ency­clopedia of Geochemistry (2000) and Capo et al. (199S). 24 CP. Bataille, GJ. Bowen j Chemical Geology 304-305 (2012) 39-52 or mantle precursors, they inherited the (87Sr/86Sr)j of their parent, but in most cases had different Rb/Sr, causing their 87Sr/86Sr to evolve along a different Rb/Sr slope than the parent material. Rocks with higher (lower) Rb/Sr than their parent evolved along a steeper (flatter) slope. Because Rb and Sr are fractionated differently in geological pro­cesses, the Rb/Sr of a rock can be modified during tectonic, metamor-phic or sedimentary transformations, thus modifying the (87Sr/86Sr)i or the slope of evolution of 87Sr/86Sr. For example, (87Sr/86Sr) and Rb/Sr values for the Idaho batholiths have been shown to vary largely along a 700 m transect depending on what host rock the batholith in­truded (King et al., 2007). Since most geological materials in the crust have been recycled multiple times throughout earth history, and this history of transformations is usually incompletely documented in geological map data, the comprehensive history of 87Sr/86Sr evolution is difficult to reconstruct. In our model, we make the simplifying assumption that the mod­ern 87Sr/86Sr of silicate rocks can be approximated based on a three stage history, where all rocks of a given lithology are assumed to have been derived from a common parent material. In each new stage, it is assumed that the new rock produced inherits its parent material's mean 87Sr/86Sr, but is differentiated chemically (Rb/Sr) from the parent material. First, 87Sr was produced in the chemically undifferentiated Earth until an 87Sr/86Sr of 0.701 was reached at 3 Ga. At 3 Ga (an approximation of the age of crustal differentiation, ti) chemical differentiation occurred, and from that time 87Sr/86Sr evolved independently in the mantle and multiple crustal rock reser­voirs. Extant rock units were formed from one of these rock reservoirs at times corresponding to their ages (t2) as documented in geological map data. This theoretical framework gives: ... (2 ) where (Rb/Sr)parent is the Rb/Sr of the parent material, and (Rb/Sr )raci< is the Rb/Sr of the modern rock. 2,2. Silicate model calibration Calculating the 87Sr/86Sr of a rock unit using Eq. (2) requires esti­mates of the parameters (Rb/Sr)parent and (Rb/Sr)raci< as well as the approximate age the rock. Information on rock age is a common fea­ture of digital geological maps, but estimating the parameters (Rb/ Sr)parent and (Rb/Sr)rock for each lithology is not as straight forward. We proceeded in two steps by calibrating the silicate model indepen­dently for each parameter. In the first calibration step, we assigned values for (Rb/Sr)parent, which determines the slope of 87Sr/86Sr evolution during stage 2. This parameter depends on the type of parent material (Fig. 1} which includes: 1) weathered bedrock for sedimentary material, 2) magma for igneous rock, 3) the parent lithology for metamorphic rocks. In the absence of information concerning the parent rock, we approximated the (Rb/Sr)parerit of sedimentary and metasedimentary rocks by assuming that they originate from a uniform source with a constant value of 0.24 corresponding to the average Rb/Sr of the upper crust (Goldstein and Jacobsen, 1988). To approximate the (Rb/Sr)parerit of igneous and metaigneous rocks, we separated these lithologies in 5 categories (ultramafic, mafic, intermediate, felsic in­termediate and felsic) using the International Union of Geological Sci­ences (IUGS) classification (Le Bas and Streckeisen, 1991). This effectively separates igneous rocks between Rb-poor mantle rocks and Rb-rich crustal rocks. We used data from the Western North American Volcanic and Intrusive Rock Database (www.navdat.org), including measurements for ( &7S r /86S r ) rtx±, a ge ( t2) and (R b /S r )rock), to back-calculate the (Rb/Sr)parent for each of 5765 samples using Eq. (2). Finally, we classified these samples according to our 5 categories and calculated the average (Rb/Sr)parent for samples in each category (see Supplementary Table 1). In the second calibration step, we assigned values to the parame­ter (Rb/Sr)rock by estimating the average Rb/Sr for 180 silicate rock unit types appearing in the United States Geological Survey state-level geological map geodatabases (Geological Survey (U.S). State Geologic Map Compilation, 2005). For each of these rock unit types, we calculated the average (Rb/Sr) from identical or analogous lithol­ogies in the USGS geochemical database (Geological Survey (U.S.) The National Geochemical Survey, 2004). We used this database because it included a large number of Rb and Sr measurements (252,661 mea­surements) covering 167 of the 180 lithologies selected. The 13 remaining lithologies were assigned (Rb/Sr)TOCk by comparison with other analogous rocks (see Supplementary Table 1). 2,3. Carbonate model calibration We modeled carbonate rocks separately because their ( 87Sr/86Sr)j is not dependent on decay, but is a function of the variations of 87Sr/ 86Sr in seawater: ^ S r \ (^Sr 86Sr , \ 86Sr . / rock \ I ) ( 3 r / rock ' (3 ) We apply the carbonate model to 10 lithologies from the USGS state-level geological map geodatabase (Supplementary material Table 1). Values of (87Sr/86Sr)seawater were estimated for each rock age (Supplementary Table 2) using 87Sr/86S r seawater curves from the Precambrian Marine Carbonate Isotope Database (PMCI) (Shields and Veizer, 2002; see Supplementary Table 3). The estimation of (Rb/Sr),-ock values for each carbonate lithology was conducted as de­scribed for silicates in Section 2.2 (see Supplementary Table 1). 2.4 Model validation We conducted separate validation exercises for igneous and sedi­mentary rocks because of the difference in calibration methods for (Rb/Sr)parent described in Section 2.2. We expect a lower accuracy of the silicate model for sedimentary rock due to the absence of infor­mation concerning the parent rock for this type of rock. We used 9130 igneous rock and 207 sedimentary rock data from the global GEOROC database (Lehnert et al., 2000). The parameterized silicate model was applied to independently predict the 87Sr/86Sr of samples represented in these databases using the parameter values from Sup­plementary Table 1 associated with the database-specified lithology, and the predicted and observed values were compared. Data from 121 samples (1.3% of the samples) were removed from the igneous rock validation dataset Among these samples, 78 (0.85% of all sam­ples) were old felsic rocks (granites or rhyolites) displaying excep­tionally high 87Sr/86Sr ranging from 0.850 to 4. These samples are also characterized by unusually high Rb/Sr ranging from 744 to 30. We recognize as a limitation of the current version of our model that it cannot accurately account for such highly radiogenic samples. The remaining 43 samples (0.47% of all samples) corresponded to rocks displaying 87Sr/86Sr values that are highly atypical for their lith­ological classification: e.g., 6 basalts were removed because their 87Sr/ 86Sr was higher than 0.730. In these cases we suspect that the data­base classifications provided an inaccurate or incomplete description of the sample lithology. We validated the carbonate model by comparing model predic­tions with 246 published data from the PMCI (Shields and Veizer, 2002) and the GEOROC database (Lehnert et al., 2000). Although this comparison does not represent a completely independent valida­tion of the model since some of the validation data were used in 25 CP. Bataille, GJ. Bowen / Chemical Geology 304-305 (2012) 39-52 reconstructing the paleo-seawater Sr isotope curves, it allows us to provide a first order assessment of model performance. Although the validation data used here provide a broad repre­sentation of lithologies and ages they are not comprehensive, and thus limit our ability to validate the model, in that: 1} analyses gathered in these databases are biased toward rocks from active tectonic and volcanic areas, 2) 87Sr/86Sr values for continental sed­imentary samples are under-represented in comparison with igne­ous rocks, and 3} Mesozoic and Cenozoic rocks represent more than 80% of the samples in the database. Additional inaccuracies in our parameterization and validation could result from a lack of control on the degree of alteration of database samples, which could lead to: 1) overestimation of Rb/Sr values because Sr is pref­erentially removed during weathering (Dasch, 1969), or 2) under­estimation of S7Sr/86Sr values because rock preferentially lose Sr from their low 87Sr/86Sr mineral phases during weathering (e.g. Bullen et al., 1996). 2.5. Mapping bedrock 87Sr/86Sr Using the above equations, we calculated Sr isotope ratios for 319,824 mapped geological units represented in the United States Geological Survey state-level geological map geodatabases (Geological Survey (U.S). State Geologic Map Compilation, 2005). Although these maps present some challenges (see Supplementary methods and http ://pubs.usgs.gov/of/2005/l 325/documents/CONUSdocumentation. pdf), they are unique in providing internally consistent, high resolution age and lithological information for the contiguous USA The 48 state lithological maps of the conterminous USA were downloaded in shape file format. Using ArcGIS, we merged the indi­vidual maps into a single shape file to obtain a geodatabase with three attributes relevant to our work: • Unit_age: the text descriptor of the maximum age of the unit, and • Rocktypel, and Rocktype2: the major and minor lithology descriptors. We used these fields to join the map unit table with a set of tables containing the parameter values used in Eqs. (2) and (3): • The table "Age" (Supplementary Table 2) listed each unique geolog­ic time descriptor found in the map units table and related the attri­bute MAXAGE with a numeric age estimated from the USGS geological time scale (Geological Survey (U.S.). Geologic Names Committee, 2007). • The tables "Lithology! " and "Lithology2" listed each lithologic descriptor present in the geodatabase and assigned values for the parameters (Rb/Sr)parent and (Rb/Sr)iithoiogy (Supplementary Table 1). • The table "Carbonates" (Supplementary Table 3) associated carbon­ate rock age with the 87Sr/86Sr of seawater. Using the values from these associated tables, we calculated 87Sr/86Sr for each map polygon (geological map unit). In cases where both major and minor lithologies were documented for a map unit, we calculated separate Sr isotope ratio estimates for each lithology. 3. W a te r m o d e ls 3. L Theory The 87Sr/86Sr of soluble Sr in stream water is largely determined by the delivery of Sr to runoff by chemical weathering of the underly­ing bedrock (Stewart et al., 1998), though in some cases the soluble Sr in water can originate from other inputs such as groundwater (Negrel and Petelet-Giraud, 2005) atmospheric deposition of sea salt and mineral dust (Chadwick et al., 2009), hydrothermal processes (Pretti and Stewart, 2002), or soils and surficial deposits (Stueber et al., 1975). Chemical weathering of bedrock is regulated by a com­plex combination of factors including lithological and mineralogical composition (Meybeck, 1987; Horton et al., 1999; Brantley et al., 2007), climate (particularly temperature and runoff; White and Blum, 1995), biology (Eckhardt, 1979; Brady and Carroll, 1994; Moulton et al., 2000) and erosion rates (Raymo et al., 1988; West et al., 2005). While these factors may be important for local studies, lithology and runoff have been identified as the two main controls of chemical weathering rates at regional scale (Beusen et al., 2009; Hartmann et al., 2009a; Jansen et al., 2010; Hartmann and Moosdorf, 2011). In order to simplify our large scale model, we limit­ed our analysis to a pair of first order lithologically-based factors influencing the flux of Sr to water: 1) differential weathering rates of rocks and minerals, which we represent as a dimensionless weath­ering rate factor W (Supplementary Table 4), and 2) differences in Sr content (Q between lithologies. The chemical weathering of carbon­ates versus silicates illustrates the importance of these factors: car­bonates have a higher Sr content and weather faster than silicates, therefore even trace quantities of calcite can be a dominant source of soluble Sr and control the 87Sr/86Sr of environmental waters (Clow et al., 1997; Anderson et al., 2000). 32. Weathering model calibration In our model, the transfer of Sr from a rock to water is given by: F = W„ormC, (4) where Fis the flux of Sr from rock to water, C is the average Sr content (Supplementary Table 4) of each rock type calculated as described for Rb/Srrock in Section 2.2 and Wnorm is the weathering rate normalized to granite (Supplementary Table 4). We adopted two different ap­proaches to estimating Wnorm, depending on rock type. To estimate Wnorm for igneous and sedimentary rocks, we calculated bulk rock dissolution rates for each rock type as: Krr (5) where i is a given mineral, U[ is abundance of f in the given rock type and K the weathering rate value for that mineral based on laboratory measurements. We estimated a[ from the IUGS classification (Le Bas and Streckeisen, 1991; Supplementary Table 4). Mineral-specific values of K were taken from averaged values of mineral weathering rates found in laboratory experiments at pH = 5.5 and T=20°C (Supplementary Table 5; Franke, 2009). However, because field stud­ies suggest that at equal mineralogical composition, relative weather­ing rates of igneous and volcanic rocks differs (Drever and Clow, 1995), we scaled our W estimate as: W (6) where is a correction factor related to differential reactive surface between rock type. Values of R were assigned by grouping rock types into three broad categories chosen to account for differences in permeability, and thus reactivity with aqueous solutions (Lewis, 1989), and comparing our calculated values of Wnorm for each catego­ry with dissolution rate measurements in small monolithic catch­ments in France (Meybeck, 1987; Meybeck, 1987). The assigned values (R=3 for volcanic rocks, R=2 for metavolcanic and R = 1 for crystalline igneous rocks) offer a rough approximation of relative field weathering rates useful for our initial large scale effort, and can be refined in future work. Because the mineralogy of metamorphic and sedimentary rocks is difficult to estimate, we estimated Wnorm of silicate sedimentary, py-roclastic and metamorphic rocks using a different approach. Based 26 CP. Bataille, G.J. Bowen / Chemical Geology 304-305 (2012) 39-52 on denudation rate measurements from small monolithic catchments (Meybeck, 1987; Meybeck, 1987), we distinguished between meta-morphic and silicate sedimentary rocks, with low weathering rates similar to those of granite (assigned Wnorm - 1) and faster-weathering argillaceous sedimentary rocks (Wnonn - 2). Because no monolithic catchment dissolution rates measurements were available for pyroclastic rocks, we estimated Wnorm from long term dissolution measurements of tuff tablets relative to those of granodiorite tablets exposed to the same conditions (Matsukura et al., 2007). Compared to regional scale estimates of chemical weathering based on dissolved silicate content (Bluth and Kump, 1994; Beusen et al., 2009; Hartmann et al., 2009; Jansen et al., 2010), our values show similar relative weathering rates across five major lithological groups: 1) car­bonates and evaporites (50>WnOrm>25)f 2) tuff, pyroclastic flow and mafic volcanic rocks (25>Wnt)rm>5), 3) other volcanic rocks and basic and intermediate igneous rocks (5>Wn0rm>3), 4) argilac-eous sediments (Wnorm -2) and 5) other metamorphic, sedimentary and felsic intrusive rock (Wnorm- 1). Although in good agreement with existing literature, our weathering formulation is limited in that it 1) does not account for runoff, climate, land cover or slope var­iations, and 2) is based on bulk dissolution rates and while accounting for differences in Sr content between rock type it does consider Sr-specific dissolution kinetics. 3.3. Mapping local and catchment water 87Srj86Sr We combined the weathering and bedrock models to map 87Sr/ 86Sr variations in local and catchment-integrated waters. The local water model estimates the &7Sr/&6Sr value of Sr leached from bedrock to water at each point on the map, whereas the catchment water model estimates the 87Sr/86Sr of surface waters flowing through each map location, including all contributions from up-catchment locations. In the local water model, for each map unit polygon where major and minor lithologies were given we calculated the relative Sr weath­ering flux from major and minor lithologies: (WC)„ = 0.75 *(WC)m - 0.25 *(WC)n O) and the average Sr isotope ratio of local water, weighted by the fluxes from major and minor lithologies: 0.75*(WC)lnQjc,r\ (WC)„t ) /0.25*(WC)min[ir\ I (WC)„ J l ssSr) (8) In these equations, the relative weights assigned to major and minor lithologies (0.75 and 0.25, respectively) represent a coarse generalization consistent with the only available constraint, that rocktypel and rocktype2 are the most and second most abundant of the rock types present in each mapped unit (e.g. http://pubs.usgs. gov/of/2005/1325/documents/CONUSdocumentation.pdf}. Sr flux and local 87Sr/&6Sr values were exported to raster data layers at 1 km spatial resolution for further analysis and mapping of catchment water 87Sr/86Sr. The catchment water map was creat­ed using the Flow Accumulation tool (Spatial Analyst toolbox) in ArcGIS and 1 km gridded flow direction values from the Hydro 1 K digital elevation model (DEM; http://edc.usgs.gov/products/ elevation/gtopo30/hydro/namerica.html). Modeled local water Sr isotope flux [(87Sr/&6Sr)jocaix (WC)tot] and Sr flux [(WC)tot] values were accumulated downstream through the DEM river networks and divided to obtain estimated water 87Sr/86Sr values that repre­sented an average of the up-stream Sr sources to each map pixel, weighted by the contribution of weathered Sr from each rock type in the catchment. We note that, although this model accounts for lithology-driven variation in weathered Sr fluxes, it does not ex­plicitly calculate the water balance of the catchment and so does not account for differences in Sr flux driven by differences in runoff from individual grid cells. 3.4. Model validation To validate the catchment water model, we compared 87Sr/86Sr predictions with water 87Sr/86Sr measured at 68 watersheds in 4 re­gions of the contiguous USA (Fig. 2). Before calculating 87Sr/86Sr in water, we obtained maps of the sub-watersheds for the Susquehanna River (http://www.srbc.net/atlas/index.asp) and the Owen Lake Basin (http://map24.epa.gov/EMR/). No pre-processed maps of the sub­watersheds existed for the Scioto River and Clark Fork of the Yellow­stone River basins. Consequently, we delineated each catchment (Fig. 2) by processing the national elevation dataset (Gesch, 2002) with the Hydrology toolbox in ArcGIS (Spatial Analyst toolbox). We successively clipped the digital elevation model (DEM) for the area considered (Geoprocessing tool/Clip Raster), reconditioned the DEM (Fill Sinks tool), calculated the flow direction (Flow Direction tool) and flow accumulation (Flow Accumulation tool) rasters, defined streams by reclassifying the flow accumulation raster (Stream Defini­tion tool; thresholds typically between 0.1 and 1% of the maximum flow accumulation), segmented the streams (Stream Segmentation tool) and finally delineated the watersheds and sub-watersheds (Wa­tershed tool). We further validated this delineation process by com­paring the shape of the catchments with the different maps furnished in published studies (Stueber et al., 1975; Fisher and Stueber, 1976; Horton et al., 1999; Pretti and Stewart, 2002). In order to test the sensitivity of the catchment water model to different modeling assumptions, we used the Spatial Statistics tool (ArcGIS Spatial Analyst toolbox) to calculate three different estimates of the average 87Sr/86Sr for each catchment: • Two estimates without weighting for differences in Sr flux among grid cells within the watershed. The first, which we call the "age-only catchment water model", is an unweighted average of 87Sr/ a6Sr values, calculated using the Beard and Johnson model (Beard and Johnson, 2000), across all grid cells in the catchment. The sec­ond, the "unweighted catchment water model" is an unweighted average of modeled ‘local water" a7Sr/86Sr values (Eq. 8) across all catchment grid cells. • A third estimate accounted for differential Sr contributions from dif­ferent map units within the catchment. This formulation, the "flux-weighted catchment water model", was equivalent to that used to map catchment water Sr isotope values as described above: the sum of the Sr isotope flux for all watershed grid cells was divided by the sum of the total Sr flux. To further test the relevance of our models for provenance appli­cations, we compared the 87Sr/86Sr predictions using these three for­mulations with the 87Sr/86Sr measured in marijuana from 79 USA counties (West et al., 2009). In this case, the samples were identified by their county of origin, and we averaged grid cell values within the county boundaries, as represented in the National Atlas of the United States (www.nationalatlas.gov/), rather than within catchments (Fig. 2). 4. R e su lts a n d d iscu s s io n 4.1. Bedrock model The silicate model explained 59% of the observed variance in an independent global dataset for 9009 igneous and metamorphic rocks and 33% of the variance for 207 sedimentary rocks (Fig. 3A and B). This new silicate model significantly improved the correlation with measurements in comparison with estimates from the age-only 27 28 29 Table 1 Geology and measured and modeled 87Sf/ssSr values for bedrock in the catchment water model validation catchments. W: watershed; CF: Clark Fork of the Yellowstone; OL: Owen Lake; Sc: Scioto; Su: Susquehanna; N = North; S = South; E = East; W = West. 46 CP. Bataille, G.J, Bowen f Chemical Geology 304-305 (2012) 39-52 w Geology Sampled lithology Measured 87Sr/86Sr Bedrock model 87Sr/86Sr Reference CF N: granitic gneiss Beartooth Mountains Granitic gneiss None 0.748 (Horton et al., 1999) S: Paleozoic marine sedimentary Andesite 0.707 Eocene Andesite Carbonates 0.7087 OL W: Sierra Nevada batholiths metavolcanic and Sierra Nevada batholiths 0.706-0.725 0.706-0.722 (Goff etal., 1991) igneous rocks E: White-Inyo mountains complex mixture of Volcanic rocks 0.706-0.708 0.707 (Goff etal., 1991) sedimentary, igneous and metamorphic rocks Tuff 0.709-0.713 0.711 (Davies and Halliday, 1998) (Marchand, 1974) Mesozoic granite 0.706-0.708 0.707 (Kistler and Peterman, 1973) Sc Mixture of Paleozoic shales and carbonates Devonian 0.7086 0.708-0.710 (Steele et al„ 1972) covered by glacial till Carbonates 0.732-0.745 0.719 Paleozoic shales 0.710 Shale leachate 0.7 OS Celestite Su N: Mixture of silicates and carbonates from Shales 0.741-0.755 0.719 (Whitney and Hurley, 1964) the Paleozoic Devonian 0.7075 0.708-0.710 limestone S: Precambrian to Mesozoic igneous, volcanic and Igneous rocks 0.707-0.799 0.707-0.752 (Wetheril et al., 1968) metamorphic rocks volcanic rocks in the West. On a regional scale (100 km), we observe large 87Sr/86Sr variations in sedimentary basins due to the difference between silicates and carbonates. High resolution 87Sr/86Sr variation is most apparent in mountainous areas due to the complex juxtaposi­tion of lithologies in these regions. Even at the scale of a county (10 km), the bedrock models (major and minor) suggest the potential for high resolution S7Sr/86Sr variations depending on the lithological complexity. 42. Catchment water models 4.2A. Water model validation The four watersheds selected for model validation represent a wide range of geological, climatic and physiographic conditions (Fig. 2, Table 1). We compiled 87Sr/86Sr measurements from 68 streams in these watersheds: 1) 13 samples from the watershed of the Clarks Fork of the Yellowstone River (WY): a mountainous catch­ment with a predominant geology of granite, andesite and carbonates (Horton et al., 1999), 2) 19 samples from the Owens Lake watershed (CA): a mountainous watershed dominated by a complex mixture of igneous and metamorphic rocks associated with dolomite (Pretti and Stewart, 2002), 3) 19 samples from the Scioto River basin (OH): a sedimentary basin dominated by shales, sandstones and marine car­bonates (Stueber et al., 1975), and 4) 18 measurements from the Sus­quehanna River basin (PA): a catchment containing varied sedimentary, metamorphic and igneous rocks (Fisher and Stueber, 1976). For each watershed the measured values reported in the liter­ature were compared with three model estimates of the catchment-integrated average water Sr isotope ratio, as described in Section 3.4. For each of these watersheds, because the bedrock models are the base maps of the catchment water model, we first tested the accuracy of the bedrock models by comparing the predictions with 87Sr/86Sr measurements of rock units within the selected watershed. Table 1 shows that for each of these watersheds, the bedrock model accurate­ly predicts the 87Sr/86Sr of most lithologies, with the exception of the lower-than-predicted 87Sr/86Sr measured for shales from OH and PA and some igneous rocks from the Wissahickon formation (Owens Lake). For the Clarks Fork of the Yellowstone Basin (Fig. 7A) each of the model formulations reproduces the basic pattern of 87Sr/86Sr differ­ences across the sub-watersheds. However, the correlation (Fig. 7B) is closer to the 1:1 relationship for the flux-weighted catchment water model than for the unweighted and age-only catchment water models. In this basin, where Cenozoic sedimentary and volca­nic rock coexists with Precambrian felsic rocks (Table 1), most of the 87Sr/86Sr variations in water are driven by the large differences in age of the different geological formations. In this geological setting, even if the age-only bedrock model does not account for differences in lithology it can be used to predict the first order patterns of varia­tion in stream water 87Sr/86Sr values with reasonable accuracy. Pre­diction accuracy was further enhanced by incorporating lithological factors (Fig. 7B). For the Owens Lake River Basin, in spite of the geological complexity of this watershed (Table 1), the 87Sr/86Sr of most streams was relatively constant at -0.710 (Fig. 7C). In most of the streams, the flux-weighted catchment water model gives a more accurate prediction and stronger correlation (Fig. 7D) than the age-only and unweighted catchment water models. Silver Creek, located in the White-Inyo Mountains, is not correctly predicted by any of the water models. This stream runs through Cambrian marine sediments, gneisses and schists (Pretti and Stewart, 2002). In the catchment water models, the 87Sr/86Sr value is buffered to low values by the presence of dolomite with a predicted 87Sr/86Sr value of 0.709 (using our bedrock carbonate calculation). However, Pretti and Stewart (2002) argued that these dolomites prob­ably exchanged Rb with shales during metamorphism and have sig­nificantly higher 87Sr/86Sr than otherwise expected, explaining the high 87Sr/86Sr in stream water of these catchments and the diver­gence with the modeled values. McGee and Convict creeks, located in the Northern part of the Basin, lack metamorphosed dolomites in outcrop, but are also poorly predicted by the flux-weighted and unweighted catchment water models. These sub-watersheds present a complex hydro-geological setting, including Paleozoic or Precambrian metasedimentary rocks which are are poorly represented by the lithological maps (Stevens and Greene, 1999). 87Sr/86Sr signature is also slightly overestimated by our models in streams within the Long Valley caldera such as Fig. 7. Catchment water model validation results. Modeled and measured 87Sr/86Sr in (A and B) 12 streams of the Clarks Fork of the Yellowstone River Basin in Wyoming (Horton et al., 1999); (C and D) 19 streams of the Owen Lake in California (Pretti and Stewart, 2002); (E and F) 19 samples from the Scioto River Basin in Ohio (Stueber et al., 1975); (G and H) IS streams of the Susquehanna River Basin in Pennsylvannia (Fisher and Stueber, 1976). Black circle: observations; Open circle: age-only water model; Open triangle: unweighted catchment water model; Reversed black triangle: flux weighed catchment water model; Black square: celestite-corrected flux-weighted catchment water model for the Scioto River Basin. Dashed lines in the right hand panels show the 1:1 relationship. 30 I9P0UJJS0B/JS 31 CP. Bataille, GJ, Boweti / Chemical Geology 304-305 (2012) 39-52 Independence and Hot creeks. This area is characterized by Low &7Sr/&6Sr rocks and hydrothermal springs which contribute greatly to the water chemistry (Pretti and Stewart, 2002). Measured 87Sr/ 86Sr values for hot springs of the area ranged from 0.7078 to 0.7081 (Goff et al., 1991) which may explain the discrepancy with our modeled 87Sr/86Sr. Pretti and Stewart (2002) also showed that these hot springs exert a strong influence at large scale, on the downstream 87Sr/86Sr value of the Owens River because of their high dissolved Sr load, an influence that would not be accounted for in our model. Other potential factors explaining the inaccuracy of the models include inputs of Sr from mineral dust (Clow et al., 1997), the poor representation of the geological complexity of these watersheds by 2D maps, or the inaccuracy of our weathering equations when several lithological weathering rates have to be ap­proximated concomitantly. For the Scioto River Basin, the unweighted catchment water model drastically overestimates the 87Sr/86Sr value in stream water in several catchments, whereas the age-only model underestimates the observed values (Fig. 7E). The flux-weighted catchment water model reduces the magnitude of overestimates relative to the unweighted catchment water model for almost all the watersheds, a difference that can be attributed to the higher Sr flux from carbonate units in the watersheds in comparison with silicates (Table 1). Several sub-watersheds in the Southern part of the Basin that are dominated by shales, such as Bear, Salt, Crooked and Scioto Brush Creeks, display low &7Sr/86Sr measurements relative to the high 87Sr/86Sr of their bedrock. This anomaly is due to the presence of minor amount of cal-cite in these shales (Table 1), which weathers preferentially and buffers the 87Sr/86Sr of the weathering flux. The Sr isotope ratio in two other sub-watersheds (Big Walnut and Scioto 3) with bedrock geology exclusively composed of silicates (Table 1) is overestimated by the flux-weighted catchment water model. However, these sub­watersheds are proximal to outcropping carbonate formations, and drillings from these catchments show that thick layers of carbonates are present at depth (Stueber et al., 1975). Groundwater discharged from these beds probably buffers the 87Sr/86Sr and explains the dis­crepancy between model and measurements (Fig. 7E). Apart from these specific examples, the catchment water model shows a general tendency to overestimate the 87Sr/86Sr of watersheds of this Basin. Stueber et al. (1975) showed that glacial overburden within the Scioto River Basin, which contains a large amount of solu­ble celestite (SrS04) and pulverized Paleozoic carbonates with 87Sr/ 86Sr equal to 0.708 (Table 1), buffers the 87Sr/86Sr in these streams (Stueber et al., 1975). To attempt to account for this factor, we devel­oped a correction for the contribution of till and carbonates to water. We based this correction on the work of Steele et al. (1972) who used the Sr concentration in water to estimate the contribution of each source of Sr. We used a surficial geology map (Clawges et al., 1999) to identify the distribution of glacial till the Basin. In each subwa­tershed covered by thick and thin glacial deposits, respectively, we considered that celestite contributed 75% (average contribution for watersheds covered by thick glacial till in Steele et al., 1972) and 50% (average contribution for watersheds covered by thin glacial till in Steele et al., 1972) of the Sr in water. The celestite-corrected water model substantially improved the accuracy of predictions with­in this Basin, explaining 81% of the variance with a model/data slope close to 1 (Fig. 7E and F). This result suggests that future work should include improved model formulations representing surficial deposits, particularly in area where thick glacial and eolian deposits are pre­sent. One impediment to this work is the relatively limited availabil­ity of systematic information on the age, origin and composition of these surficial deposits. For the Susquehanna River Basin, the flux-weighted catchment water model dramatically improves the model predictions relative to the age-only and unweighted catchment water models (Fig. 7G and H). In most of the watersheds, the unweighted catchment water model overestimates the 87Sr/86Sr whereas the flux-weighted catchment water model matches the observations closely because of the importance given to preferential dissolution of carbonates. The improved performance of the flux-weighted catchment water model is seen here despite the divergence between bedrock model predic­tions and measurements for shales (Table 1). Similar to the Scioto Basin, shale leachates here have a significantly lower 87Sr/86Sr than the whole rock due to the selected dissolution of minor amount of calcite (Table 1). The discrepancy between the flux-weighted catch­ment water model and observed 87Sr/86Sr values for Deer and Octor-aro creeks can be explained by the inability of the bedrock model to accurately predict the 87Sr/86Sr of rocks from the Wissahickon Forma­tion (Table 1). 422. A global view of the catchment water model The flux-weighted catchment water model explains 70% of the variance of the Sr isotopes in water for the 68 watersheds tested with a linear correlation close to the 1 : 'l relationship (Fig. 8). Predic­tion accuracy for this model, estimated based on the validation data, is significantly improved relative to the other models, with MAE = 0.00051 and RMSE = 0.0034. In comparison, the age-only catchment water model explains 38% of the observed variance with MAE= -0.0039 and RMSE = 0.0056. In our approach, we added a number of lithological effects that increased the accuracy of water 87Sr/86Sr predictions in most of the geological settings. The resulting local water (Fig. 9A) and flux-weighted catchment water (Fig. 9B) maps for the contiguous USA show patterned 87Sr/86Sr variations similar to the bedrock models (Fig. 6A and B). Average 87Sr/86Sr values are highest in the new bedrock model and lowest in the flux-weighted catchment water model, where the preferential dissolution of low 87Sr/86Sr units (e.g. carbonates and mafic rocks) buffers the 87Sr/86Sr of the water catchment model in comparison with bedrock (Fig. 6D). The maps predict large variations at a range of spatial resolutions, which are promising for provenance studies. Nevertheless, the 87Sr/ 86Sr prediction in water could be improved by considering the poten­tial contribution of non-bedrock sources of Sr to water (Sillen et al., 1998; Stewart et al., 1998). In our validation process, we demonstrat­ed the importance of accounting for the contribution of Sr-rich min­erals (calcite, dolomite and celestite) because they often buffer the 87Sr/86Sr of whole rivers. Similarly, the effect of dust deposition in the Rockies (Clow et al., 1997), contributions from soil and surficial materials (Stewart et al., 1998) and the effect of local phenomenon such as hydrothermal contributions (Pretti and Stewart, 2002) and atmospheric deposition (Stewart et al., 1998) should be considered in future work. 0.75 0.74 s __E 0.73 in *>-. 0.72 inCO 0.71 0.70 age-only model ▼ flux weighted w a ter model y = 1.05 (x) - 0.033 r* = 0.70 y = 0.38 (x) + 0.44 r* = 0.38 0.70 0.71 0.72 0.73 0.74 S7 86 Srobservatlon 0.75 Fig. 8, Validation of the catchment water Sr isotope model across all study catchments, showing linear regressions between measured S7Sr/86Sr and flux-weighted catchment water and age-on |y water model predictions for 68 streams of the USA (celestite-corrected values are used for the Scioto River). Dashed line shows the 1:1 relationship. 32 33 CP. Bataiile, CJ. Bowen / Chemical Geology 304-305 (2012) 39-52 0.730 _ 0.725 TJ 0.720 CO CO 0.715 W CO 0.710 0.705 0.700 A / / / / V T / ▼ s T V t a ▼ 4 y = 0 ,6 6 x+ 0 ,2 4 / 1 * * 0 4 5 B / / A / A / A / A p . A A £ / / y=0.52x+0.34 r*=0-13 0.73 0.70 o/ / <D y=0,21x+0,55 r*=0 29 observation Kg. 10. Linear regression between S7Sr/86Sr in marijuana and the mean values of the modeled water in the county of sample origin. W ater values are the average of all grid cells in the county, weighted as described for the (A) flux-weighted catchment water model, (B) unweighted catchment water model, and (C) age-only catchment water model. Symbols as in Fig. 7. Dashed line in each panel shows the 1:1 relationship. formulation represents major lithology-specific effects and yet re­mains generalized to the extent that it could be applied in any region where basic digital geological map data (including lithology and age} are available. In spite of the limitations discussed throughout this paper, this new mapping method represents a significant advance in modeling major environmental Sr sources to ecosystems, and the strength of the correlations between the different models and the ob­servations are encouraging. Moreover, although the predictive power of the model remains limited in many cases, our documentation of model performance through quantitative comparisons with observa­tional data allows informed use of the model-derived data products. A number of regions of the contiguous USA display promising 87Sr/86Sr variations at different scales which could be used to deter­mine rock, water or biological material provenance. The Sr isoscapes could complement other existing isoscapes (Bowen et al., 2005} used for provenance studies because 87Sr/86Sr varies widely at re­gional and continental scales. The development of more detailed and harmonized seamless geological maps for the conterminous USA (Jansen et aL, 2010} and other regions, as well as refined high resolution lithological studies and geochemical sampling, could rap­idly improve the resolution and accuracy of these isoscapes. In this respect, we suggest the following as critical next steps to improving the predictability of environmental Sr isotope ratios at large scales: 1} Develop more flexible parameterizations and parameter distribu­tions that increase the ability of the model to represent highly ra­diogenic rock units. 2) Improve weathering rate calculations by including functions de­scribing rate dependence on factors such as runoff, climate, pedol­ogy, vegetation and topography. Recent work from Jansen et al. (2010} offers a good starting point. 3) Develop submodels representing the contribution of atmospheric sources of Sr, particularly dust and sea salt, to soil and bioavailable Sr. 4} Identify systematic approaches to representing the contribution of Sr weathered from surficial deposits to water. 5) Ultimately, model and scale the contribution of these sources of Sr, including bedrock weathering, atmospheric sources and surficial Sr sources, to bioavailable and biological pools of interest, includ­ing soil, soil water, surface and groundwater and organismal Sr. The work of Stewart et al. (1998) provides a platform on which such an effort could be developed. 6) In all cases, these efforts will be advanced through the continued accumulation and compilation of Sr isotope measurements and el­emental concentration data from a range of materials. Our results suggest that the most critical data gaps vary depending on location, but include measurements of surficial deposits that are not adequately characterized on most geological maps and contin­ued analysis of Sr isotope ratios of stream systems, which inte­grate the geochemistry of their watersheds and can reveal important inaccuracies in the model. A c k n ow led g em en ts This research was supported by NSF Award DBI-0743543. We thank Casey Kennedy and Hilary Bataiile for reviewing and comment­ing on this manuscript, and three anonymous reviewers for their comments which greatly improved this manuscript. A p p en d ix A. S u p p lem e n ta ry d a ta Supplementary data to this article can be found online at doi: 10.1016/j.chemgeo.2012.01.028. Refe ren ces Anderson, S.R, Drever, J.I., Frost, CD., Holden, P., 2000. Chemical weathering in the foreland of a retreating glacier. Geochimica et Cosmochimica Acta 64 (7), 1173-1189. Barnett-John son, R., Pearson, T.E., Ramos, F.C., Grimes, CB., Mac Far lane, R.B., 2008. Tracking natal origins of salmon using isotopes, otoliths, and landscape geology. Limnology and Oceanography 53 (4), 1633-1642. Beard, B.L., Johnson, CM., 2000. Strontium isotope composition of skeletal material can determine the birth place and geographic mobility of humans and animals. Journal of Forensic Sciences 45 (5), 1049-1061. Bentley, R.A, 2006. Strontium isotopes from the earth to the archaeological skeleton: a review. Journal of Archaeological Method and Theory 13 (3), 135-187. doi: 10.1007/s10816-006-9009-x. Bentley, RA., Knipper, C, 2005. Geographical patterns in biologically available stron­tium, carbon and oxygen isotope signatures in prehistoric SW Germany. Archaeo­metry 47, 629-644. Bentley, R.A., Wahl, J., Price, T.D., Atkinson, T.C., 2008. Isotopic signatures and heredi­tary traits: snapshot of a Neolithic community in Germany. Antiquity 82 (316), 290-304. Beusen, A.H.W., Bouwman, A.F., Durr, H.H., Dekkers, A.LM., Hartmann, J., 2009. Global patterns of dissolved silica export to the coastal zone: results from a spatially ex­plicit global model. Global Biogeochemical Cycles 23. Bluth, G.J.S., Kump, L.R., 1994. Lithologic and climatologic controls of river chemistry. Geochimica e t Cosmochimica Acta 58 (10), 2341-2359. Bowen, G.J., 2010. I so scapes: spatial pattern in isotopic biogeochemistry. Annual Re­view of Earth and Planetary Sciences 38, 161-187. doi: 10.1146/annurev-earth- 040809-152429. Bowen, G.J., Wassenaar, LI., Hobson, K.A., 2005. Global application of stable hydrogen and oxygen isotopes to wildlife forensics. Oecologia 143 (3), 337-348. doi: 10.1007/s00442-004-1813-y. Brady, P.V., Carroll, S.A, 1994. Direct effects of C02 and temperature on silicate weath­ering - possible implications for climate control. Geochimica et Cosmochimica Acta 58 (7), 1853-1856. Brantley, S.L., Goldhaber, M.B., Ragnarsdottir, K.V., 2007. Crossing disciplines and scales to understand the Critical Zone. Elements 3 ( 5), 307-314. 34 Cft Bataille, G.J. Bowen / Chemical Geology 304-305 (2012) 39-52 51 Balien, T.D., Krabbenhoft, D.P., Kendall, C., 1996. Kinetic and mineralogic controls on the evolution of groundwater chemistiy and 87Sr/86Sr in a sandy silicate aquifer, northern Wisconsin, USA. Geochi mica et Cosmochimica Acta 60 (10), 1807-1821. Capo, RC., Stewart, B.W., Chadwick, OA, 1998. Strontium isotopes as tracers of ecosys­tem processes: theory and methods. Geoderma 82 (1-3), 197-225. Carlson, R.W., 2003. Introduction to Volume 2. In: Heinrich, D.H., Karl, K.T. (Eds.), Treatise on Geochemistry. Pergamon, Oxford, pp. xv-xxiii. Chadwick, Q.A., Derry, L.A, Bern, C, 2009. Changing sources of strontium to soils and ecosystems across the Hawaiian Islands, Chemical Geology 267, 64-76. doi: 10.1016/j.chemgeo.2009.01.009. Chamberlain, C.P., et al., 1997. The use of isotope tracers for identifying populations of migratory birds. Oecologia 109 (1), 132-141. Chaudhuri, S., Clauer, N., 1993. Strontium iso topic compositions and potassium and ru­bidium contents of formation waters in sedimentary basins - clues to the origin of the solutes. Geochimica et Cosmochimica Acta 57 (2), 429-437. Clawges, R.M., Price, C.V., National Water-Quality Assessment Program (U.S.) and Geological Survey (U.S), 1999. Digital data sets describing principal aquifers, sur-ficial geology, and ground-water regions of the conterminous United States. U S Geological Survey open-file report 99-77. USGS, Rapid City, S.D. Clow, D.W., Mast, M.A., Bullen, T.D., Turk, J.T., 1997. Strontium 87 strontium 86 as a tracer of mineral weathering reactions and calcium sources in an alpine/subalpine watershed, Loch Vale, Colorado. Water Resources Research 33 (6), 1335-1351. Crittenden, R.G., et al., 2007. Determining the geographic origin of milk in Australasia using multi-element stable isotope ratio analysis. International Dairy Journal 17 (5), 421-428. doi: 10.1016/j.idairyj.2006.05.012. Dasch, E.J., 1969. Strontium isotopes in weathering profiles, deep-sea sediments, and sedimentary rocks. Geochimica et Cosmochimica Acta 33 (12) 1521-&. Davies, G.R., Halliday, A.N., 1998. Development of the Long Valley rhyolitic magma system: strontium and neodymium isotope evidence from glasses and individual phenocrysts. Geochimica et Cosmochimica Acta 62 (21-22), 3561-3574. DePaolo, D.J., 1981. Trace element and isotopic effects of combined wall rock assimila­tion and fractional crystallization. Earth and Planetary Science Letters 53 (2), 189-202. doi: 10.1016/0012-82lx (81)90153-9. Drever, J.I., Clow, D.W., 1995. Weathering rates in catchments. In: Brantley, S., White, A. (Eds.), Chemical weathering rates of silicate minerals. : Rev. in Mineral., vol. 31. Mineral. Soc. of Am., Washington, D. C, pp. 463-483. Eckhardt, F.E.W., 1979. Uber die Einwirkung heterotropher Mikroorganismen auf die Zersetzung silikatischer Minerale. Zeitschrift fur Pflanzenernahrung und Boden-kunde 142 (3), 434-445. Ezzo, J.A, Johnson, CM., Price, T.D., 1997. Analytical perspectives on prehistoric migration: a case study from east-central Arizona. Journal of Archaeological Science 24 (5), 447-466. Faure, G., 1977. Principles of isotope geology. Related Information: Smith and Wyllie intermediate geology series. Medium: X; Size: Pages: 475 pp. Fisher, R.S., Stueber, A.M., 1976. Strontium isotopes in selected streams within Susque­hanna River Basin. Water Resources Research 12 (5), 1061-1068. Franke, W.A., 2009. The durability of rocks - developing a test of rock resistance to chemical weathering. American Journal of Science 309 (8), 711-730. doi: 10.2475/ 08.2009.04. Frei, R., Frei, K.M., 2011. The geographic distribution of strontium isotopes in Danish surface waters - a base for provenance studies in archaeology, hydrology and agriculture. Applied Geochemistry 26, 326-340. Geological Survey (U.S.), 2004. The National Geochemical Survey, database and documentation, U S Geological Survey open-file report 2004-1001. U.S. Dept, of the Interior, U.S. Geological Survey, [Reston, Va.|. Geological Survey (U.S), 2005. State Geologic Map Compilation, U S Geological Survey open-file report (various), online at http://tin.er.usgs.gov/geology/state/. Geological Survey (U.S.). Geologic Names Committee, 2007. Divisions of geologic time - major chronostratigraphic and geo chronologic units. Fact sheet 2007-3015. US. Geological Survey, [Reston, Va.]. Gesch, D., 2002. The National Elevation Dataset. Photogrammetric Engineering and Remote Sensing 68 (1) 5 -+ . Goff, F., Wollenberg, H.A., Brookins, D.C, Kistler, R.W., 1991. A Sr-isotopic comparison between thermal waters, rocks, and hydrothermal calcites, Long Valley Caldera, California. Journal of Volcanology and Geothermal Research 48 (3-4), 265-281. Goldstein, S,J„ Jacobsen, S.B., 1988. Nd and Sr isotopic systematics of river water sus­pended material - implications for crustal evolution. Earth and Planetary Science Letters 87 (3), 249-265. Graustein, W.C, 1989. S7Sr/S6Sr ratios measure the sources and flow of strontium in terrestrial ecosystems. Stable isotopes in ecological research. Ecological Studies 68,491-512. Graustein, W.C., Armstrong, R.L, 1983. The use of Sr-87 Sr-86 ratios to measure atmo­spheric transport into forested watersheds. Science 219 (4582), 289-292. Grousset, F.E., Biscaye, P.E., 2005. Continental aerosols, isotopic fingerprints of sources and atmospheric transport: a review. Chemical Geology 222, 149-167. Hartmann, J., Moosdorf, N., 2011. Chemical weathering rates of silicate-dominated lithological classes and associated liberation rates of phosphorus on the Japanese Archipelago - implications for global scale analysis. Chemical Geology, doi: 10.1016/j.chemgeo.2010.12.004. Hartmann, J., Jansen, N., Durr, H.H., Harashima, A, Okubo, K., Kempe, S., 2009. Predict­ing riverine dissolved silica fluxes into coastal zones from a hyperactive region and analysis of their first order controls. International Journal of Earth Sciences, do i: 10.1007/S00531 -008 -0381 -5. Hodell, D.A, Quinn, RL, Brenner, M., Kamenov, G., 2004. Spatial variation of strontium isotopes (Sr-87/Sr-86) in the Maya region: a tool for tracking ancient human mi­gration. Journal of Archaeological Science 31 (5), 585-601. doi:10.1016/j.jas. 2003.10.009. Hogan, J.F., Blum, J.D., Siegel, D.I., Glaser, P.H., 2000. Sr-87/Sr-86 as a tracer of ground­water discharge and precipitation recharge in the Glacial Lake Agassiz Peat lands, northern Minnesota. Water Resources Research 36 (12), 3701-3710. Hoppe, K.A, Koch, P.L., Carlson, R.W., Webb, S.D., 1999. Tracking mammoths and mastodons: reconstruction of migratory behavior using strontium isotope ratios. Geology 27 (5), 439-442. Horton, T.W., Chamberlain, C.P., Fan tie, M., Blum, J.D., 1999. Chemical weathering and lithologic controls of w ater chemistry in a high-elevation river system: Clark's Fork of the Yellowstone River, Wyoming and Montana. W ater Resources Research 35 (5), 1643-1655. Jansen, N., Hartmann, J., Lauerwald, R., Durr, H.H., Kempe, S., Loos, S., Middelkoop, H., 2010. Dissolved silica mobilization in the conterminous USA. Chemical Geology 270 (1-4), 90-109. doi: 10.1016/j.chemgeo.2009.11.008. Kawasaki, A, Oda, H, Hirata, T, 2002. Determination of strontium isotope ratio of brown rice for estimating its provenance. Soil Science and Plant Nutrition 48 (5), 635-640. Kelly, S., Heaton, K., Hoogewerff, J., 2005. Tracing the geographical origin of food: the application of multi-element and multi-isotope analysis. Trends in Food Science & Technology 16 (12), 555-567. doi:10.1016/j.tifs.2005.08.008. King, E.M., Beard, B.L., Valley, J.W., 2007. Strontium and oxygen iso topic evidence for strike/slip movement of accreted terranes in the Idaho Batholith. Lithos 96, 387-401. Kistler, R.W., Peterman, Z.E., 1973. Variations in Sr, Rb, K, Na, and initial Sr-87-Sr-86 in Mesozoic granitic rocks and intruded wall rocks in Central California. Geological Society of America Bulletin 84 (11), 3489-3511. Koch, P.L., et al., 1995. Isotopic tracking of change in diet and habitat use in African elephants. Science 267 (5202), 1340-1343. Le Bas, M.J., Streckeisen, A.L., 1991. The IUGS systematics of igneous rocks. Journal of the Geological Society 148 (5), 825-833. doi: 10.1144/gsjgs.l48.5.0825. Lehnert, K., Su, Y., Langmuir, CH., Sarbas, B., No hi, U., 2000. A global geochemical data­base structure for rocks. Geochemistry, Geophysics, Geosystems 1 (5), 1-14. doi: 10.102 9/199 9gc000026. ' " Lewis, M.A, 1989. Water. In: McCall, J., Marker, B. (Eds.), Earth Science Mapping for planning, development and conservation. Graham and Trotman. Marchand, D.E., 1974. Chemical weathering, soil development, and geochemical frac­tionation in a part of the White Mountains, Mono and Inyo Counties, California. Geological Survey Professional Paper 352-J: Erosion and Sedimentation in a Semi-arid Environment. U.S. Govt. Print. Off, Washington, pp. 379-424. v p. Matsukura, Y., Hattanji, T., Oguchi, C.T., Hirose, T., 2007. Ten year measurements of weathering rates of rock tablets on a forested hills lope in a humid temperate region, Japan. Zeitschrift Fur Geomorphologie 51 (1), 27-40. doi: 10.1127/0372­8854/ 2007/005 ls-0027. Meybeck M., 1987. Global chemical weathering of surficial rocks estimated from river dissolved loads. American Journal of Science 287 (5), 401-428. Miller, E.K., Blum, J.D., Fried land, AJ„ 1993. Determination of soil exchange able-cation loss and weathering rates using Sr isotopes. Nature 362 ( 6419), 438-441. Moulton, K