Assessing Pyrite-Derived Sulfate in the Mississippi River with Four Assessing Pyrite-Derived Sulfate in the Mississippi River with Four Years of Sulfur and Triple-Oxygen Isotope Data Years of Sulfur and Triple-Oxygen Isotope Data

11 Riverine dissolved sulfate (SO 42− ) 12 sulfur and oxygen isotope variations 13 reflect their controls such as SO 42− 14 reduction and re-oxidation, and source 15 mixing. However, unconstrained temporal 16 variability of riverine SO 42− isotope 17 compositions due to short sampling 18 durations may lead to mischaracterization 19 of SO 42− sources, particularly for the 20 pyrite-derived sulfate load. We measured 21 the sulfur and triple-oxygen isotopes (δ 34 S, 22 δ 18 O, and ∆’ 17 O) of Mississippi River SO 42− with biweekly sampling between 2009-2013 to test 23 isotopic variability and constrain sources. Sulfate δ 34 S and δ 18 O ranged from −6.3‰ to −0.2‰ 24 and −3.6‰ to +8.8‰, respectively. Our sampling period captured the most severe flooding and 25 drought in the Mississippi River basin since 1927 and 1956, respectively, and a first year of 26 sampling that was unrepresentative of long-term average SO 42− . The δ 34 S SO4 data indicate pyrite- 27 derived SO 42− sources are 74 ±10% of the Mississippi River sulfate budget. Furthermore, pyrite 28 oxidation is implicated as the dominant process supplying SO 42− to the Mississippi River, 29 whereas the ∆’ 17 O SO4 data shows 18 ±9% of oxygen in this sulfate is sourced from air O 2 .

This is the accepted peer-reviewed manuscript (postprint) of "Assessing pyrite-derived sulfate in the Mississippi River with four years of sulfur and triple-oxygen isotope data" by Bryan Killingsworth (bryan.a.killingsworth(at)gmail.com), Huiming Bao, and Issaku Kohl The final, copy-edited, version was published in Environmental Science & Technology and can be found at https://doi.org/10. 1021/acs.est.7b05792 This postprint was provided by the publisher as a courtesy for the fulfillment of funder open access requirements and is submitted to EarthArXiv by the authors.
14 reduction and re-oxidation, and source 15 mixing.

Introduction 34
The characterization of riverine chemical fluxes today is important for establishing 35 natural baselines for understanding the magnitude of human impact on chemical cycles and 36 interpreting the rock record of biogeochemical changes in Earth's past. For the surface sulfur 37 cycle, which is closely linked to the carbon and oxygen cycles over long time scales, its most 38 significant flux is riverine sulfate input to the ocean 1 . Natural variations in the magnitude and 39 isotopic composition of the sulfate flux reflects the exposures and weathering rates of sulfide and 40 sulfate minerals in rocks 2 . Meanwhile, human activities, such as the mining and burning of fossil 41 fuels like coal, can increase riverine sulfate fluxes four-fold and alter sulfate's sulfur isotope 42 composition on a large scale, such as in the Mississippi River 3 . Thus, studies of the continental 43 sulfur cycle have made efforts to constrain riverine SO 4 2− fluxes and isotope compositions 44 globally 4-7 and in large 8-10 and small [11][12][13][14] rivers. Riverine studies have suggested that global 45 sulfate budgets may underrepresent the pyrite-derived sulfate flux, which is particularly 46 important for sulfur and carbon weathering budgets 9, 15 . 47 An advantage of riverine chemical studies is their integration of spatial and temporal 48 scales. Depending on the research focus, sampling campaigns of different durations can be 49 designed according to the basin size and its assumed variability, to constrain temporal variation 50 with multi-year sampling, or to reveal spatial variations. However, a lack of temporal constraints 51 on SO 4 2− isotopes may result in biased conclusions about sulfate sources and processes. For 52 example, the average isotope compositions of inputs are needed to construct stable isotope 53 mixing models to constrain sulfate sources in individual rivers and in models of the global 54 surface sulfur cycle. A low frequency or short duration of sampling may bias model results 55 towards one season or an anomalous year. Only some riverine sulfate isotope studies last a year 8, 56 10,[16][17][18] or longer [19][20][21][22][23][24][25][26] , and out of the rivers that were monitored for ≥1 year, just the Yangtze, 57 Indus, Oldman, and Kalix are >5,000 km 2 . For these long-term studies, the average ranges for 58 δ 34 S and δ 18 O are ~5‰ and have no correlation with catchment size. Thus there are insufficient 59 temporal constraints on riverine SO 4 2− isotope variability over large, and continental, spatial 60 scales. For example, significant variability was shown within one year of Yangtze River sulfate 61 data where the ranges of δ 34 S SO4 and δ 18 O SO4 were respectively 9.5‰ and 8.9‰ 10 . In another 62 example from a highly cited study, 85% ±5% of sulfate flux in the relatively pristine Mackenzie 63 River in Canada was attributed to pyrite oxidation 9 . This assessment was based on stable isotope 64 (δ 34 S and δ 18 O) mixing of sulfate sources for 20 samples taken throughout the basin at one time, 65 with only one sample recovered from the river mouth that could represent the output to the 66 ocean. The respective δ 34 S and δ 18 O had significant spatial variation of 28.3‰ and 12.6‰, 67 respectively, and the temporal variability was undetermined. While the conclusion appears 68 robust that most of the sulfate in the Mackenzie is pyrite-derived, it remains difficult to know if 69 the estimate of pyrite-derived sulfate flux is applicable to the long term. Regardless, the 70 estimation of pyrite-derived sulfate loads, and its natural and anthropogenic partitioning, is not a 71 trivial task due to large δ 34 S and δ 18  respectively. We note that the linear "capital delta" definition is 121 to the standard VSMOW (Fig. S1). The choice of the reference slope C in equation (3) (Table S2). with reported values between +16.7‰ to −34.7‰ but mostly negative with a mean at −19.7‰ 40 . 204 The implication for respectively lower and higher δ 34 S SO4 of the Missouri River and Ohio River 205 is consistent with the seasonal and spatial patterns in δ 34 S from zebra mussels across the 206 Mississippi River basin during 1997-1998 41 . The zebra mussel sulfur is sourced from riverine 207 sulfate but slightly fractionated, for example at Baton Rouge the reported mussel δ 34 S was near 208 −4‰ versus the average δ 34 S SO4 around −3‰ from our study, with their comparison suggesting 209 that Mississippi River sulfate end members have not changed significantly over the past ~20 210 years. 211 2 1 2

Mass balance of pyrite-derived sulfate 213
Here we provide a more detailed Mississippi River sulfate mass balance that expands on 214 the previous work using the δ 34 S SO4 average 3 . The δ 34 S of sulfate from rock weathering of shale 215 pyrite and evaporite (excluding mine drainage), which here we will call δ 34 S RW , was previously 216 estimated as −6.5‰, where the sulfate from natural and anthropogenically enhanced rock 217 weathering were assumed to have the same δ 34 S value 3 . We adopt δ 34 S values of 20‰ for 218 evaporite 9 and −17‰ for pyrite from marine shales 42 , for δ 34 S E and δ 34 S Py respectively. Despite a 219 typically wide range in pyrite sulfur isotope compositions, such as a ~50‰ range in Cretaceous 220 pyrite δ 34 S data 40 , the strong δ 34 S difference between average marine evaporite and pyrite is 221 forgiving when using estimated values to partition their mass balance. With the given 222 constraints, the mixing equation becomes: 223 The mass balance of pyrite and evaporite can then be solved for the unknown fraction of pyrite, 225 f Py , which is determined as 0.72 with evaporite being the remainder. Thus, the sulfate load from 226 rock weathering in the Mississippi River is 72% pyrite-derived sulfate and 28% evaporite sulfate, 227 or a respective 26.5% and 11% of the total Mississippi River sulfate budget (Table S3) (Fig. 2). Mississippi River sub-basins 50 and this should register in the Mississippi River that has 319 significant pyrite-derived sulfate that takes most of its oxygen from water. However, water data 320 for the Ohio and Upper Mississippi 59 , and Missouri 60 Rivers, indicates respective average 321 δ 18 O water for these rivers of −7.5‰, −8.2‰, and −9‰. The average oxygen isotope differences 322 are a maximum of 1.5‰ between Mississippi River sub-basin river waters and this will be 323 reflected in their pyrite-derived sulfate. Thus, while the δ 34 S SO4 response is strong, the δ 18 O SO4 324 responds weakly to the geographic origin of sulfate sources in the Mississippi River (Fig. 3b). 325    (Fig. 4). A "hydrological drought" condition is where surface and ground water 340 availability is lower than average due to meteorological drought, as caused by anomalously low 341 precipitation that can in turn be caused by temperature anomalies 62 . The PHDI can be considered 342 an indicator of environmental response to precipitation input, and as such the PHDI changes 343 more slowly than precipitation 63 . The strong correlation between Mississippi River δ 18 O SO4 and 344 PHDI, but lack of correlations between δ 18 O SO4 and changes in sulfate flux between Mississippi 345 River sub-basins or other chemical parameters (e.g., USGS-monitored concentration of redox-346 sensitive elements such as As and V, water temperature, and dissolved oxygen) suggests that it is 347 a balance between sulfate from more recent surface runoff versus sulfate from groundwater that 348 controls Mississippi River δ 18 O SO4 . It is possible that oxygen source for low δ 18 O SO4 could be 349 from the northwestern region of the Mississippi River basin, as streamwaters in the upper 350 Missouri River area can range down to a δ 18 O H2O of −18‰ 50 . However, Mississippi River sulfate 351 should also show the low δ 34 S SO4 values expected from Missouri River sulfate input if it was 352 more significant during drought but this does not occur (Fig. 2) In experiments 53 , and in natural systems 65 , it has been observed that the oxidation of pyrite under 374 submersed and alternating wet/dry conditions results in sulfate with δ 18 O SO4 that is around 2‰ to 375 18‰ higher than that of ambient water δ 18 O water , a scenario that was used to estimate the ranges 376 Variations in ∆' 17 O SO4 can help to differentiate sources, formation pathways, and 390 processes affecting riverine sulfate. However, ∆' 17 O SO4 in surface waters is an underdeveloped 391 tracer, with presently only one riverine ∆ 17 O SO4 study available 66 . Here we note that ∆ 17 O (eq. 2) 392 is used to describe triple oxygen isotopes in general, but we report the logarithmic form ∆' 17 O 393 (eq. 3). The difference between ∆' 17 O and ∆ 17 O is very small for measurements not far from the 394 origin, for example within error of each other for the Mississippi River sulfate data (Fig. S1).  composition has an almost entirely water oxygen source by simply checking its closeness to 441 meteoric water in triple oxygen isotope space (Fig. 5). We assume that the highest Mississippi 442 River ∆' 17 O SO4 (−0.01‰, Fig. 5 concentrations that were interpolated, from measurements taken approximately monthly, to give 477 daily values. Then, the ratios for sub-basin sulfate fluxes were each sub-basin's sulfate flux 478 versus their combined sum, here using the mix of sulfate flux between three sub-basins to 479 represent the whole Mississippi River. The difference in sulfate flux between the averages 480 determined from the lower Mississippi River and the summed three sub-basins was 6% during 481 the study period, and thus the flux contribution from the middle Mississippi River is neglected in 482 the model. The influence of hydrological conditions represented by the Palmer Hydrological 483 Drought Index is split into "wet" and "dry" components in order to assign different respective 484 "wet" and "dry" isotope values (δ xx Z Wet and δ xx Z Dry ). The PHDI f Wet and f Dry components use 485 reported monthly PHDI interpolated to give daily PHDI, which is then scaled to make a ratio 486 where maximum dry PHDI during the study period is equal to 1, and wet PHDI is the difference, 487 where  with measured data. Measured data error bars are shown or are smaller than symbols. The flux-508 only model uses a mix of sulfate flux from three Mississippi River sub-basins, the PHDI-only 509 model simulates forcing from overall hydrological conditions for the contiguous United States, 510 and the mixed model incorporates both flux-only and PHDI-only models with further details and 511 discussion given in the main text. 512 5 1 3 514

Implications for riverine sulfate 515
Mississippi River ∆' 17 O SO4 , δ 18 O SO4 , and δ 34 S SO4 each reveal their own different 516 perspectives on the Mississippi River system and its response to seasonal changes or year-to-year 517 weather patterns. Although our Mississippi River study reveals characteristics of riverine sulfate 518 that might be widespread, each river should be considered as a more-or-less unique case, with its 519 own set of processes and sulfur sources dictated by climate, hydrology, and geology. Moreover, 520 in other rivers, anthropogenic influence on sulfate may be expressed by isotopic shifts in the 521 opposite direction, to lower δ 34 S SO4 for example, as compared to what is inferred from the 522 Mississippi River. Although not done for this study, measurement of the triple oxygen isotope 523 composition (δ 18 O and ∆' 17 O) of not only the dissolved sulfate, but also the river water from the 524 same sample, could enable high resolution sulfate oxygen isotope mass balance calculations, 525 further assist in tracing river water sources such as runoff versus groundwater, and aid 526 interpretations of the sources of sulfate oxygen and sulfate oxidation pathways. Prime targets for 527 follow-up sulfate sulfur and oxygen isotope studies would be the Missouri and Ohio river sub-528 basins to characterize the loadings of pyrite-derived sulfate from natural and anthropogenic 529 bedrock weathering and mine drainage. Our results add to the calls for reassessing the 530 contribution of pyrite-derived sulfate to global sulfur budgets 9 , especially pyrite-derived sulfate 531 from coal mining 15 , and suggest that the important estimates of natural and anthropogenic global 532 riverine sulfate flux 5, 73 are due for an update. 533 5 3 4 Acknowledgements 535 Thanks to Bill Alvey and Tell City High School in Indiana for Ohio River water sampling, Justin 536 Hayles, Stefan Lalonde, Manuel Bellanger, Pierre Sans-Jofre, and the members of Bao group 537 past and present for helpful discussions, and the anonymous reviewers whose constructive 538 comments greatly improved this manuscript. National Science Foundation (EAR-1251824, EAR-539 1312284 to HB) provided part of the research fund, and indirect support was received during the 540 preparation of this manuscript from the European Union Horizon 2020 research and innovation 541 program (Marie Sklodowska-Curie grant agreement No 708117 to BK). 542 5 4 3

Supporting Information Available 544
Supporting Information includes additional details on methods, modeling, oxygen isotope 545 discussion, and tables of sulfate isotope data, an updated Mississippi River sulfate budget, and 546 model inputs. 547