Variability and Trends in the Tropical Pacific and the El Niño-Southern Oscillation Inferred from Coral and Lake Archives

Home , Onotoa

Variability and trends in the tropical Pacific and the El Niño-Southern Oscillation inferred from coral and lake archives

Item Type text; Electronic Dissertation

Authors Thompson, Diane Marie

Publisher The University of Arizona.

Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.

Download date 05/10/2021 09:00:42

Link to Item http://hdl.handle.net/10150/311122

VARIABILITY AND TRENDS IN THE TROPICAL PACIFIC AND THE EL NIÑO- SOUTHERN OSCILLATION INFERRED FROM CORAL AND LAKE ARCHIVES

by Diane Marie Thompson ______A Dissertation Submitted to the Faculty of the DEPARTMENT OF GEOSCIENCES In Partial Fulfillment of the Requirements For the Degree of DOCTOR OF PHILOSOPHY In the Graduate College THE UNIVERSITY OF ARIZONA

2013

THE UNIVERSITY OF ARIZONA GRADUATE COLLEGE

As members of the Dissertation Committee, we certify that we have read the dissertation prepared by Diane M. Thompson entitled “Variability and trends in the tropical Pacific and the El Niño-Southern Oscillation inferred from coral and lake archives” and recommend that it be accepted as fulfilling the dissertation requirement for the Degree of Doctor of Philosophy

______Date: Julia Cole

______Date: Jonathan T. Overpeck

______Date: Joellen Russell

______Date: Warren Beck

______Date: Alexander W. Tudhope

Final approval and acceptance of this dissertation is contingent upon the candidate’s submission of the final copies of the dissertation to the Graduate College.

I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement.

______Date: Dissertation Director: Julia Cole

STATEMENT BY AUTHOR

This dissertation has been submitted in partial fulfillment of requirements for an advanced degree at the University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.

Brief quotations from this dissertation are allowable without special permission, provided that accurate acknowledgment of source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his or her judgment the proposed use of the material is in the interests of scholarship. In all other instances, however, permission must be obtained from the author.

SIGNED: Diane M. Thompson 4

ACKNOWLEDEMENTS

This dissertation would not have been possible without the help and support of many individuals. I would like to thank my advisor, Julia Cole, and committee members

Jonathan Overpeck, Joellen Russell, Warren Beck and Sandy Tudhope for their guidance and feedback on my research. I am also grateful for the contributions of my collaborators

Michael Evans, Julien Emile-Geay, Toby Ault, Jessica Conroy, David Noone, Suz

Tolwinski-Ward, Kevin Anchukaitis, Mark Bush, and Aaron Collins. The field work necessary for this research would not have been possible without assistance from Sarah

Truebe, Mandy Miller, Noemi Dozouville, Roberto Pepolas, Diego Ruiz, Alexander

Tudhope, Meriwether Wilson, and Colin Chilcot. I’m forever grateful to Heidi Barnett,

Stephan Hlohowskyj and David Steinke for their hard work assuring I could generate data from these samples. I’d like to thank fellow graduate students and colleagues for stimulating science discussions and support: Jessica Conroy, Nick McKay, Sarah Truebe,

Toby Ault, Cody Routson, Sarah White, Dan Griffin, Adam Csank, Steph McAfee,

Kendra Murray, and many others. Finally, I’d like to thank family and friends for their support, particularly my parents Charles and Marilyn Thompson and my sister Susie.

This research was supported by several funding sources: an Institute for the Study of Planet Earth (ISPE) Scholarship, a Wilson Thompson Scholarship, a Susan G. Earl

Galileo Circle Endowed Scholarship, an Institute of the Environment Dissertation

Improvement Grant, a Paul S. Martin Scholarship, a Philanthropic Education

Organization (PEO) Scholar Award, a National Oceanic and Atmospheric Administration grant, and a NSF Rapid grant. 5

DEDICATION

I dedicate this dissertation to my father, Charles Thompson, who cultivated and encouraged my love of science and the natural world.

TABLE OF CONTENTS

LIST OF FIGURES ...... 9 LIST OF TABLES...... 12 ABSTRACT...... 13 1. INTRODUCTION ...... 16 2. PRESENT STUDY...... 23 3. REFERENCES ...... 27 4. APPENDICES ...... 32

APPENDIX A: COMPARISON OF OBSERVED AND SIMULATED TROPICAL CLIMATE TRENDS USING A FORWARD MODEL OF CORAL 18O...... 32 Abstract……………………………………………………………………………..…34 1. Introduction……………………………………………………………………...…35 2. Forward model of coral δ18O………………………………………………………37 3. Input datasets………………………………………………………………………38 4. Comparison of observed and pseudocoral networks………………………………38 5. Results……………………………………………………………………………..40 5.1 Pseudocorals derived from instrumental observations…………………………..40 5.2 Pseudocorals derived from CGCM output…………………………….…………42 6. Discussion……………………………………………………………………….…43 7. Conclusions and future work…………………………………………………...... 44 Acknowledgements………………………………...…………………………………47 Reference list…….……………………………………...……….……………………47 S1. Supplementary methods……………………………………..……………………55 S1.1. Forward modeled ‘pseudocoral’- observed coral comparisons………………...55 S1.2. Mean state and ENSO-related variance…………………………………….…..56 Supplemental references………………………………………………………………64 APPENDIX B: ENHANCED E-W TEMPERATURE GRADIENT ACROSS THE DATE LINE OVER THE 20th CENTURY INFERRED FROM CENTRAL PACIFIC CORAL RECORDS……………………………………………………………………..67 Abstract…………………………………………………………………………..……68 1. Introduction…………………………………………………………………..……69 2. Study area…………………………………………………………………….……73 3. Methods……………………………………………………………………...…….74 3.1 Coral sampling…………………………………………………………...……74 3.2 Geochemical analysis………………………………………………………….75 3.2.1 Elemental analysis…………………………………………………….………75 3.2.2 Stable isotope analysis………………………………………………………...76 3.3 Chronology development……………………………………………………...76 3.4 Sr/Ca-SST calibration…………………………………………………………77 18 3.5  Osw reconstruction…………………………………………..……..………79 3.6 Uncertainty estimation………………………………………………….....….80 7

4. Results……………………………………………………………………………..83 4.1 Replication of coral Sr/Ca and variations…………………………...….83 4.2 Coral Sr/Ca-SST reconstruction…………………………….……………….85

4.3 Coral , Sr/Ca-SST and variability……………………………..86 4.4 20th-century trends………………………………………………...…………86 5. Discussion……………………………………………...………………………….89 5.1 Reproducibility of coral Sr/Ca………………………………………...……89 5.2 Sr/Ca-SST relationship……………………………………………………...91

5.3 Coral , Sr/Ca-SST and variability…………………………….92

5.4 Coral-derived trends in tropical Pacific SST, and gradients……….94 5.4.1 Coral-derived SST trends…………….………………………………….94

5.4.2 Coral-derived trends……………………………………………97 6. Conclusions…………………………………………………………………….…99 Acknowledgements…………..……………………………………………………....101 References…………………………………………………………………………....102

APPENDIX C: 20th-CENTURY VARIABILITY IN TRADE WIND REVERSALS INFERRED FROM Mn/Ca IN A WESTERN PACIFIC CORAL……………………..143 Abstract………………………………………………………………………………144 1. Introduction………………………………………………………………………145 2. Tarawa coral record……………………………………………………………...147 3. Zonal wind……………………………………………………………………….148 4. Results……………………………………………………………………………150 4.1 A new record of El Niño-related westerly winds………………………..………150 4.2 20th-century westerly wind trends……………………………………………….152 5. Discussion…………………………………………………………..……………153 6. Conclusions and future work…………………………………………………….155 Acknowledgements……………………………………………………………..……156 Reference list…………………………………………………………………...……156

APPENDIX D: CLIMATE CONTROLS ON BAINBRIDGE CRATER LAKE, GALÁPAGOS OVER THE LAST 6000 YEARS………….……………………….…167 Abstract………………………………………………………………………………168 Introduction………………………………………………………………………….169 Bainbridge Crater Lake………………………………………………………..……172 Materials and Methods………………………………………………………………173 Environmental monitoring………………………………………………………….173 Sediment sampling………………………………………………………………….175 Age modeling…………………………………………………………………...…..178 Results……………………………………………………………………..…………180 Climate and limnology……………………………………………………………...180 Water chemistry………………………………………………………………….…184 Sedimentation………………………………………………………………..….….185 8

Age model…………………………………………………………...... 188 Climate history from Bainbridge Sediment cores……………………………….…189 Discussion……………………………………………………………………………191 Conclusions…………………………………………………………….………….…196 Acknowledgements………………………………………………………………….197 References………………………………………………………………………...…197

APPENDIX E: PERMISSIONS………………………………………...……...………233 Permission for Appendix A from John Wiley and Sons…….………………………..234

LIST OF FIGURES Figure 1. Tropical Pacific mean SSTs and El Niño anomalies………………………..…19 Figure A-1. Magnitude of the trend slope………………………………………………..52

Figure A-2. Change in mean state (x-axis) and ENSO-related variability………………53

Figure A-3. Correlation of SST and SSS and SSS trend fields……………………….…54

Figure A-S1. Spatial and temporal pattern of observed coral ENSO and trend PC…..…62

Figure A-S2. Correlation of SST and SSS and SSS trends…………………………...…63

Figure B-1. Mean SSTs, El Niño anomalies and site locations……………………..….115

Figure B-2. Positive X-ray images of Onotoa coral cores…………………………...…116

Figure B-3. Positive X-ray images of the Maiana 95-2-3 coral core……………..….…117

Figure B-4. Positive X-ray images of the Jarvis 99-1-2 coral core……………...……..118

Figure B-5. SEM images of Jarvis 2 surfaces with and without alteration…………….119

Figure B-6. Sensitivity of the Onotoa Sr/Ca-SST trend to the calibration………..……120

Figure B-7. Sensitivity of the Maiana Sr/Ca-SST trend to the calibration……………..121

Figure B-8. Sensitivity of the Jarvis 2 Sr/Ca-SST trend to the calibration…………..…122

Figure B-9. Onotoa and Maiana 18O, Sr/Ca-SST, and 18Osw and trends…..………..123

Figure B-10. Jarvis 1 & 2 18O, Sr/Ca-SST, and 18Osw and trends…………...……..123

Figure B-11. Sr/Ca-SST trends and E-W SST gradient trend…………………...……..124

Figure B-12. Sr/Ca-SST trends and N-S SST gradient trend………………………...…125

Figure B-13. Sr/Ca-SST trends and E-W SST gradient trend using mean calibration…126

Figure B-14. Sr/Ca-SST trends and N-S SST gradient trend using mean calibration….127

Figure B-15. Sr/Ca-SST trends and E-W SST gradient trend 1900-1998……………...128

Figure B-16. Sr/Ca-SST trends and N-S SST gradient trend 1900-1998……………....129

18 Figure B-17. Salinity and  Osw trends over the 1972-1992 period…………….……..130 10

18 Figure B-18. Sensitivity of the Onotoa  Osw trend to the calibration………………...131

18 Figure B-19. Sensitivity of the Maiana  Osw trend to the calibration……………..….132

18 Figure B-20. Sensitivity of the Jarvis 2  Osw trend to the calibration……………..….133

Figure B-21. Patterns of the decadal variability…………………………………….….134

Figure B-22. Interannual Sr/Ca-SST variability and E-W variability gradient…….…..135

Figure C-1. Tarawa atoll δ18O and Mn/Ca records and zonal wind reanalysis………..163

Figure C-2. Mn/Ca and zonal wind variance and westerly wind frequency………..….163

Figure C-S1. Zonal wind anomalies and number of observations………………..……164

Figure D-1. Bainbridge study region, aerial image, and bathymetric map……………..210

Figure D-2. Local climate and limnology of Bainbridge Lake 12/2009-10/2012…...…211

Figure D-3. Histograms of seasonal weather station data………………………….…..212

Figure D-4. Histograms of seasonal sonde data………………………………………..213

Figure D-5. Histograms of warm season weather station data……………………...….214

Figure D-6. Histograms of warm season sonde data……………………..……….……215

Figure D-7. Histograms of weather station data by sediment trap period…………..….216

Figure D-8. Histograms of sonde data by sediment trap period…………………….….217

Figure D-9. Images and elemental maps rock and sediment samples……………….…218

Figure D-10. Boxplot of Al/Si ratio in rock and soil samples…………………….……219

Figure D-11. Images of sediment trap samples from 1 and 2m depth……………...….220

Figure D-12. Histogram of calcium intensity in sediment trap samples………………..221

Figure D-13. BSE images and WDS spectra of sediment and rock samples………..…222

Figure D-14. Log-log scatterplot of Mg/Mn vs Na/K in rock, sediment and dust…….223

Figure D-15. Age models for the Bainbridge 2007, 2009, and 1991 cores……………224

Figure D-16. The Bainbridge sediment record and radiocarbon ages…………………225 11

Figure D-17. Image and XRF maps of elemental intensities of core Bain 2B……..….226

Figure D-18. Bainbridge sediment and regional records since 2500 cal years BP……227

Figure D-19. Bainbridge sediment and regional records for the past 6000 cal years.…228

Figure D-S1. Comparison of sonde and water sample data………………………….…230

Figure D-S2. Sensitivity test of the Bainbridge 1991 age model……………………...231

LIST OF TABLES

18 Table A-S1. Slope of the δ Osw-salinity relationship……………………………...……59

Table A-S2. Regression statistics…………………………………………….………….60

Table A-S3. Root mean square error statistics…………………………………………..61

Table B-1. Coral site description………………………………………………….……136

Table B-2. Correlation between coral geochemical records……………………………137

Table B-3. Mean and 20th century trends in coral geochemical records…………….…138

Table B-4. Correlation between coral records and local SST………………………….139

Table B-5. Correlation between coral records and regional SST indices………………140

Table B-6. Correlation between SST gradient indices and regional SST indices…...…141

Table B-7. Regression between coral records and regional SST indices………………142

Table C-S1. Strength of historical El Niño events in Tarawa wind records……………165

Table D-1. Bainbridge YSI profile June 2010………………………………………….205

Table D-2. Bainbridge water chemistry in 2009, 2010, and 2012………………...……206

Table D-3. Radiocarbon measurements and calibrated ages………………….…..……208

Table D-4. Number of carbonate and siliciclastic laminae in the sediment record…….209

ABSTRACT The background state and changes associated with the El Niño-Southern

Oscillation (ENSO) in the tropical Pacific Ocean influence climate patterns all over the world. Understanding how the tropical Pacific will be impacted by climate change is therefore critical to accurate regional climate projections. However, sparse historical data coverage and strong natural variability in the basin make it difficult to assess the response of the tropical Pacific to anthropogenic climate change. Further, climate models disagree regarding the response of the basin to continued anthropogenic forcing into the future.

Building off of the limited instrumental record, high-resolution records from coral and lake sediment archives can be used to assess the response of the tropical Pacific to past climate changes and to compare and assess climate model projections. In the present study, I use high-resolution coral and lake records from the equatorial Pacific to assess climate model projections and the response of the coupled ocean-atmospheric climate system in the basin (ocean temperature, salinity, winds, precipitation) to natural and anthropogenic forcing.

Using a simple model of how climate is recorded by corals, we compare historical climate data and climate model simulations with coral paleoclimate records to assess climate model projections and address uncertainties in the historical data, models and paleoclimate records. We demonstrate that this simple model is able to capture variability and trend observed in the coral records, and show that the both sea surface temperature and salinity contribute to the observed coral trend. However, we find major 14

discrepancies in the observed and climate model simulated trends in the tropical Pacific that may be attributed to uncertainties in model simulated salinity.

We then assess 20th-century variability and trends in SST and salinity in the central tropical Pacific using replicated coral δ18O and Sr/Ca records from the Republic of Kiribati and the central Line Islands. We find that the coral records from these sites display a warming and freshening trend superimposed on strong interannual and low- frequency variability. Further, we demonstrate an apparent strengthening of the E-W

SST gradient across the dateline (between 173°E and 160°W) and a slight weakening of the N-S SST gradient due to enhanced warming along the equator and west of the dateline relative to other sites. However, we find no evidence of increased variability in the central Pacific, suggesting that there has not been an increase in central Pacific style

ENSO events. Finally, we show that the salinity response to climate change may be very patchy within the basin.

Using a new ~90 year coral Mn/Ca record from the central Pacific, we investigate variability and trends in tropical Pacific trade winds. First, we demonstrate a strong association between westerly wind anomalies and coral skeletal Mn/Ca, which recorded all of the major historical El Niño events of the 20th century. In this new long Mn/Ca record, we find a reduction in the amplitude and frequency of Mn/Ca pulses between

1893 and 1982, suggesting a decrease in westerly wind anomalies in the western equatorial Pacific Ocean. 15

Finally, we use a sediment record from Bainbridge Crater Lake, Galápagos

Archipelago to assess variability in the eastern tropical Pacific over the past ~6 thousand years. Based on results from long-term monitoring of the lake, we propose a new climate interpretation of the sediment record and find further evidence reduced mid-Holocene

ENSO variability and a ramp up of ENSO variability starting around 1775 cal. years BP. 16

INTRODUCTION

1.1 Significance

Understanding the effects of climate change on the El Nino-Southern Oscillation

(ENSO) is crucial for regional climate predictions, because ENSO events drive intense floods, droughts and temperature anomalies all over the world. However, despite the importance of ENSO to climate on a global scale, observational and model products disagree regarding the trends in the tropical Pacific (Vecchi et al. 2008); therefore, it is still unclear how the tropical Pacific Ocean and ENSO will be impacted by climate change. Determining the expected range of future ENSO-related variability is especially important here in the Southwest, where water resources are particularly vulnerable and strong ENSO-related droughts exacerbate the already depleted water supply. The network of high-resolution coral and lake records from across the equatorial Pacific developed in this work provide insight into the response of this system to past changes in radiative forcing, and may thus help constrain predictions for future change within the tropical Pacific.

1.2 Background

1.2.1 - Variability within the tropical Pacific and hypotheses for the future

Approximately every 3-7 years, the strong easterly trade winds within the tropical

Pacific Ocean weaken and anomalously warm temperatures stretch across the basin, heralding an El Niño event (e.g., Figure 1). Because these temperature anomalies alter atmospheric circulation patterns far outside the Pacific, El Niño events (and their cold 17

water counterparts, La Niña events) affect climate worldwide, driving intense floods, droughts and temperature anomalies.

Two main responses to greenhouse gas (GHG) forcing over the tropical Pacific have been proposed, one in which the east-west zonal sea-surface temperature (SST) gradient strengthens and one in which the zonal gradient weakens. Due to the similarities of these patterns to those that develop during ENSO events, these responses have been referred to as ‘La Niña-like’ and ‘El Niño-like’ trends, respectively. There have been numerous mechanisms proposed to explain these two quite opposite responses within the tropical Pacific. Differential latent heat flux responses, atmospheric stability and cloud feedbacks have been invoked to explain the El Niño-like response to warming in the tropical Pacific. For instance, enhanced evaporative damping of SSTs in the western

Pacific relative to the eastern Pacific may lead to a reduced SST gradient (Wallace 1992,

Knutson and Manabe 1995). Similarly, a reduced SST gradient may result from differential shortwave and longwave cloud-cover feedbacks, with negative cloud-cover forcing over the west (Ramanathan and Collins 1991, Meehl and Washington 1996,

Meehl et al. 2000), and positive cloud-cover forcing over the east (Klein and Hartmann

1993, Meehl and Washington 1996, Meehl et al. 2000). This response may also result from a reduction in the strength of the Walker circulation and increased atmospheric stability as a result of a greater change in atmospheric water vapor content relative to precipitation under GHG forcing (Held and Soden 2006). In contrast, Clement et al.

(1996) invoke ocean dynamics to explain the response of the tropical Pacific to uniform warming, suggesting that anomalous upwelling offsets warming in the eastern equatorial 18

Pacific and results in an increased SST gradient. Model experiments incorporating natural solar and volcanic forcing changes over the past 1000 years demonstrate that heating forced by increased solar activity and/or decreased volcanic activity in the past have resulted in an increased SST gradient within the Pacific (Mann et al. 2005).

However, it is likely that the response of the tropical Pacific to GHG forcing will be quite different from its response to natural forcing over past millennia. Some suggest that the ‘ocean thermostat mechanism’ may be the transient response to uniform forcing

(Liu 1998, Seager and Murtugudde 1997), with differential latitudinal heating (with more warming in the extratropics) counterbalancing this cooling on longer timescales (Liu

1998, Seager and Murtugudde 1997). Therefore, these two proposed responses of the tropical Pacific (‘El Niño-like’ vs ‘La Niña-like’) may not be mutually exclusive, and may operate on different timescales. Furthermore, DiNezio et al. (2009) suggest that a reduction in the strength of the Walker circulation does not require a significant change in the zonal SST gradient, suggesting the ocean and atmosphere may decouple under future warming.

Changes in mean state of the tropical Pacific may in turn modulate the strength, frequency, and spatial pattern of ENSO events (e.g., Federov and Philander, 2000; Yeh et al., 2009; Li et al., 2011). Decadal changes in the strength and frequency of ENSO events, with some of the strongest events of the instrumental record during the last couple of decades, suggest that interannual variability in the tropical Pacific is closely tied to the background state (e.g., Federov and Philander, 2000). Further, the recent prevalence of

ENSO events with maximum anomalies in the central Pacific (central Pacific or El Niño 19

Modoki events, Figure 1b), suggests that the spatial pattern of ENSO events may also be influenced by the background state and greenhouse-gas forcing (e.g., Yeh et al., 2009).

However, whether such decadal variations in the strength, frequency and spatial pattern of ENSO reflect forced changes or the response to noise in the nonlinear system (e.g.,

Wittenberg 2009, Yeh et al., 2011) is still a matter of considerable debate.

Figure 1: (a) December ERSST v2 SST anomaly (ºC) averaged over the past three major El Niño events: 1972, 1982, and 1997 and (b) December 2009 ERSSTv2 SST anomalies (ºC), where the ‘x’s’ mark the locations of our samples from (west to east): Jarvis Island, the Republic of Kiribati (Maiana and Onotoa), and the Galapagos (Bainbridge Crater Lake).

1.2.2 - The balance of evidence

Climate model simulations disagree regarding trends within the tropical Pacific

(e.g., Yeh et al., 2012), making it difficult to address how the basin will respond to future 20

climate change. Although most coupled global climate models (CGCMs) and atmospheric models (with simplified oceans) suggest an El Niño-like response to GHG forcing (Collins et al. 2005, Meehl et al. 2007, Vecchi and Soden 2007, Vecchi et al.

2008, Guilyardi et al. 2012), slab ocean models with a simplified atmosphere (e.g., Cane-

Zebiak Model) suggest a La Niña-like response. Model simulations also disagree greatly regarding future variability within the tropical Pacific (Meehl et al. 2007, Guilyardi et al.

2012, Kim and Yu 2012, Stevenson et al. 2012), suggesting that the relationship between mean state and variability is still not well understood. Although there is an emerging consensus that the tropical Pacific will experience enhanced equatorial warming and a shoaling of the thermocline under future greenhouse gas forcing (Liu et al., 2005;

Guilyardi et al. 2009; Gastineau and Soden, 2009; DiNezio et al., 2009; Collins et al.

2010; Widlansky et al., 2013), multimodel comparison studies suggest that the most likely response of the tropical Pacific SST gradient will be one of no change (Collins et al. 2005, Liu et al. 2005, Guilyardi et al. 2009).

Due to the disagreement among climate model projections for the tropical Pacific, recent research has turned to historical observations to assess the response of the basin to

20th-century warming. However, historical products disagree regarding the trend in the

SST gradient (Vecchi et al. 2008, Deser et al. 2010), with the magnitude and sign of the trend dependent on the season (Karnauskas et al. 2009) and time period (Liu et al. 2005) analyzed. Further, observations that suggest the SST gradient has strengthened (Cane et al. 1997, Hansen et al. 2005, Hansen et al. 2006, Karnauskas et al. 2009, An et al., 2011) appear to be at odds with the weakening of the Walker circulation inferred from SLP 21

observations (Vecchi et al. 2006, Zhang and Song 2006, Bunge and Clarke 2009,

Tokinaga et al., 2012) given that the ocean and atmosphere are strongly coupled in the tropical Pacific (Bjerknes 1969). Further, although SLP anomalies suggest that there has been a weakening of the Walker circulation, there is no clear corresponding trend in zonal wind anomalies (e.g., Wu and Xie, 2003, McVicar et al., 2008).

Because historical climate record is too short and uncertain to determine the range of natural variability or to desern the response of the tropical Pacific Ocean to anthropogenic forcing, high-resolution proxy records from the basin are needed to determine the effect of anthropogenic forcing on the background state of the tropical

Pacific and the strength of ENSO. However, there have been few direct tests of these hypotheses using proxy records, and most that have addressed this issue have relied on the translation of the ENSO-related SST anomaly through the atmosphere, often to remote areas outside of the tropical Pacific. Due to the nonlinear nature of these atmospheric teleconnections, the relationship between ENSO and these proxy records may change through time, muddling their interpretation. Further, although long lake and tree-ring records provide a means to assess millennial to centennial variability in the tropical Pacific, their annual at best resolution is typically too low to decipher changes in

ENSO-related variance from changes mean state or determine how these trends may vary seasonally. As many of these proxy records may only record ENSO-related anomalies in a certain season, the integration of the climate signal in these low-resolution proxies may also result in a seasonal bias. Although typically much shorter in length, records from annually banded corals provide a means to obtain sub-annually resolved records of 22

surface ocean conditions at the time of their growth, and may thus provide direct proxy records for past changes in the tropical Pacific. In this work, we capitalize on the advantages of high-resolution records from both corals and lakes at sites across the equatorial tropical Pacific (Figure 1) to assess changes in the background state and ENSO variability in response to changes in natural and anthropogenic forcing.

PRESENT STUDY

Presented here are four studies that utilize climate model simulations and a multiproxy network of paleoclimate records from across the equatorial tropical Pacific

Ocean to address changes in mean state and ENSO variability within the basin. Each publication is presented in an appendix and is formatted as required by the specific journals. All articles are coauthored, and I am senior author on all four manuscripts.

In the first Appendix (A) we use a simple model of how climate is recorded in the coral skeleton to convert historical climate data and climate model simulations into synthetic coral records (“pseudocorals”). In doing so, this simple model allows us to not only assess climate model projections for the tropical Pacific, but also address uncertainties in the historical climate data, climate model simulations and the coral paleoclimate records. Modeling pseudocorals with historical data, we demonstrate that this simple model is able to capture the variability and trend observed in the network of coral records over the 20th century. We then use the model to assess the magnitude of the observed coral trend and the relative contribution of sea surface temperature and salinity to this trend, important remaining questions in coral paleoclimatology. Finally, we compare pseudocorals modeled from historical simulations of coupled general circulation models with the observed coral network. We find major discrepancies in the observed and simulated trends in the tropical Pacific, and find that these discrepancies may be attributed (at least in part) to uncertainties in the observed and modeled salinity field. 24

These results emphasize the need for additional proxy reconstructions of salinity and

18 δ Osw (Appendix B).

In Appendix B we use replicated records from the Republic of Kiribati and the central Line Islands to assess 20th-century variability and trends in SST and salinity in the central tropical Pacific. We use these new records, along with the previously published record from Palmyra Atoll (6ºN, 162ºW , [Nurhati et al. 2009; 2011a], to address the spatial fingerprint of warming within the basin in terms of changes in the E-W SST gradient across the dateline (between 173°E and 160°W) and the N-S equatorial SST gradient. We also capitalize on the differential impact of eastern Pacific and central

Pacific type ENSO events at these sites to address trends in these two types of ENSO events. First, we demonstrate that Sr/Ca-SST variations are highly reproducible between two coral records from Kiribati, despite being separated by over 400 km. We find that

the Sr/Ca-SST, , and records from these sites suggest a warming and freshening trend superimposed on strong interannual and low-frequency variations related to ENSO and the modes of Pacific decadal SST variability. Further, the coral records suggest that the warming at Kiribati (173-175°E) may have been greater than further east at Jarvis (160°W), resulting in an apparent strengthening of the SST gradient.

In contrast, the warming trend is similar in magnitude between Jarvis near the equator, and Palmyra at 6ºN, suggesting that there has not been a strong enhancement of warming along the equator or change in the N-S SST gradient. We also find no detectable increase

CP variability associated with CP type ENSO events over the 20th century. Finally, we 25

show that coral reconstructions suggest that the salinity response to climate change may be very patchy within the basin.

In the third Appendix (C) we investigate variability and trends in tropical Pacific trade winds to assess the response of the Walker circulation to warming. Despite the importance of trade wind strength and direction as fundamental components of ENSO and the Walker circulation, historical observations of tropical Pacific winds are limited and disagree on long-term trends. In this work, we use a new ~90 year coral Mn/Ca record from Tarawa Atoll and zonal wind data from the 20th century reanalysis project to demonstrate a strong association between westerly wind anomalies and coral skeletal

Mn/Ca. This ~90 year coral Mn/Ca record captures the major historical El Niño events of the 20th century, including several strong events that are underestimated or absent in historical SST and wind datasets. This record therefore provides the first long reconstruction of trade winds in the ENSO active region of the tropical Pacific. We find a reduction in Mn/Ca variability resuting from a decrease in the amplitude and frequency of Mn/Ca pulses within this 90-year record, suggesting a decrease in westerly wind anomalies in the western equatorial Pacific Ocean between 1893 and 1982.

In the final Appendix (D), we use a sediment record from Bainbridge Crater Lake,

Galápagos Archipelago to put the 20th-century changes into the context of variability observed over the past ca. six thousand years. It has been proposed that the kind of laminations preserved in the Bainbridge sediment record may provide a record of El Niño events of different intensities (Riedinger et al. 2002); however, current hypotheses for how the ENSO events are preserved in the lake sediment record have not been directly 26

tested. In this work, we monitored the local climate and limnology of the lake to determine how seasonal to interannual climate variability gets recorded in the sediment record. Based on results from long-term monitoring, we propose a new climate interpretation of the sediment record. The Bainbridge sediment record provides support for a reduced mid-Holocene ENSO (between ~6100-4000 cal. years BP), followed by a stepwise increase in ENSO variability starting around 2500 cal. years BP. Taken together, the Bainbridge sediment record and other available ENSO reconstructions from the tropical Pacific basin suggest that the tropical Pacific zonal SST gradient weakened and ENSO-related variability strengthened around 1775 (± 190) cal. years BP. The

Bainbridge sediment record suggests that ENSO variability remained high until ~900 years BP, when sedimentation rate slowed and ENSO variability at the lake decreased as the tropical Pacific zonal SST gradient strengthened. Over the past 900 years, the sediment record suggests that La Niña-like conditions dominated between 400-600 BP and between ~350 BP and 0 BP (1950) when gypsum precipitated in the lake, and wetter,

El Niño-like conditions dominated after 0 BP (1950). The results of this study suggest a strong connection between the strength of the zonal SST gradient and ENSO-related variability within the tropical Pacific.

REFERENCES An, S.-I., J.-W. Kim, S.-H. Im, B.-M. Kim, and J.-H. Park, 2011. Recent and future sea surface temperature trends in the tropical Pacific warm pool and cold tongue regions. Climate Dynamics, doi:10.1007/s00382-011-1129-7.

Bjerknes, J. 1969. Atmospheric teleconnections from the equatorial pacific 1. Monthly Weather Review 97(3): 163-172.

Bunge, L., and A.J. Clarke, 2009. A Verified Estimation of the El Niño Index Niño-3.4 since 1877. J. Clim. 22:3979-3992.

Cane MA, Clement AC, Kaplan A, Kushnir Y, Pozdnyakov D, Seager R, Zebiak SE, and Murtugudde R, 1997. Twentieth-Century Sea Surface Temperature Trends. Science 275:957-960

Clement AC, Seager R, Cane MA, and Zebiak SE, 1996. An Ocean Dynamical Thermostat. Journal of Climate 9:2190-2196

Collins M, The CMIP Modelling Groups, 2005. El Niño- or La Niña-like climate change? Climate Dynamics 24:89-104

Collins, M., An, S. I., Cai, W., Ganachaud, A., Guilyardi, E., Jin, F. F., ... & Wittenberg, A. (2010). The impact of global warming on the tropical Pacific Ocean and El Niño. Nature Geoscience, 3(6), 391-397.

Deser, C., A.S. Phillips, and M.A. Alexander (2010), Twentieth century tropical sea surface temperature trends revisited. Geophysical Research Letters, 37(10).

DiNezio, P.N., A.C. Clement, G.A. Vecchi, B.J. Soden, B.P. Kirtman, and S.-K. Lee, 2009. Climate Response of the Equatorial Pacific to Global Warming, J. Clim.22: 4873-4892. 28

Fedorov A.V., and Philander S.G., 2000. Is El Niño Changing? Science 288:1997-2002.

Gastineau, G., and B. J. Soden, 2009. Model projected changes of extreme wind events in response to global warming. Geophysical Research Letters 36.

Guilyardi, E., A. Wittenberg, A. Fedorov, M. Collins, C. Wang, A. Capotondi, G.J. van Oldenborgh, and T. Stockdale (2009), Understanding El Niño in ocean-atmosphere general circulation models: progress and challenges, Bull. Amer. Meteorol. Soc., 90(3), 325-339.

Guilyardi, E., H. Bellenger, M. Collins, S. Ferrett, W. Cai, and A. Wittenberg (2012), A first look at ENSO in CMIP5, Clivar Exchanges, 17(1), 29-32.

Hansen, J., L. Nazarenko, R. Ruedy, M. Sato, J. Willis, A. Del Genio, D. Koch, A. Lacis, K. Lo, S. Menon, T. Novakov, J. Perlwitz, G. Russell, G.A. Schmidt, and N. Tausnev, 2005. Earth’s Energy Imbalance: Confirmation and Implications. Science 308:1431- 1435.

Hansen, J., M. Sato, R. Ruedy, K. Lo, D.W. Lea, and M. Medina-Elizade, 2006. Global temperature change. PNAS 103(39):14288-14293.

Held, I.M., B.J. Soden, 2006. Robust Responses of the Hydrological Cycle to Global Warming. J. Clim. 19:5686-5699.

Karnauskas, K.B., R. Seager, A. Kaplan, Y. Kushnir, and M.A. Cane, 2009. Observed Strengthening of the Zonal Sea Surface Temperature Gradient across the Equatorial Pacific Ocean. J. Clim. 22: 4316-4321.

Kim, S. T., and J.-Y. Yu 2012. The two types of ENSO in CMIP5 models, Geophys. Res. Lett. 39: L11704, doi:10.1029/2012GL052006 29

Klein, S.A., and D.L. Hartmann, 1993. The Seasonal Cycle of Low Stratiform Clouds. J. Clim. 6:1587-1606.

Knutson, T.R., and S. Manabe, 1995. Time-Mean Response over the Tropical Pacific to Increased CO2 in a Coupled Ocean-Atmosphere Model. J. Clim. 8:2181-2199.

Li, J. B., et al., 2011c. Interdecadal modulation of El Nino amplitude during the past millennium. Nature Climate Change 1: 114-118. Liu, Z., 1998. The Role of Ocean in the Response of Tropical Climatology to Global Warming: The West-East SST Contrast. J. Clim. 11:864-875. Liu, Z., S. Vavrus, F. He, N. Wen, and Y. Zhong, 2005. Rethinking Tropical Ocean Response to Global Warming: The Enhanced Equatorial Warming. J. Clim. 18:4684- 4700.

McVicar, T. R., et al. (2008), Wind speed climatology and trends for Australia, 1975– 2006: Capturing the stilling phenomenon and comparison with near-surface reanalysis output. Geophysical Research Letters, 35(20), L20403.

Mann ME, Cane MA, Zebiak SE, and A Clement. 2005. Volcanic and Solar Forcing of the Tropical Pacific over the Past 1000 Years. Journal of Climate 18:447-456

Meehl, G.A., and W.M. Washington, 1996. El Niño-like climate change in a model with increased atmospheric CO2 concentrations. Nature 382:56-60.

Meehl, G.A., W.D. Collins, B.A. Boville, J.T. Kiehl, T.M.L. Widley, and J.M. Arblaster, 2000. Response of the NCAR Climate System Model to Increased CO2 and the Role of Physical Processes. J. Clim. 13:1879-1898.

Meehl, G.A., Stocker, T.F., Collins, W.D., Friedlingstein, A.T., Gaye, A.T., Gregory, J.M., Kitoh, A., Knutti, R., Murphy, J.M., Noda, A., Raper, S.C.B., Watterson, I.G., Weaver, A.J. and Zhao, Z. (2007) Global Climate Projections. In: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth 30

Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, pp. 747-845. ISBN 9780521880091.

Ramanathan, V., and W. Collins, 1991. Thermodynamic regulation of ocean warming by cirrus clouds deduced from observations of the 1987 El Niño. Nature 351:27-32.

Riedinger, M. A., M. Steinitz-Kannan, W. M. Last, and M. Brenner (2002), A similar to 6100 C-14 yr record of El Niño activity from the Galapagos Islands, J Paleolimnol, 27(1), 1-7.

Seager, R., and R. Murtugudde, 1997. Ocean Dynamics, Thermocline Adjustment, and Regulation of Tropical SSTs. J. Clim 10:521-534. Stevenson, S., B. Fox-Kemper, M. Jochum, R. Neale, C. Deser, and G. Meehl, 2012. Will there be a significant change to El Nino in the twenty-first century? Journal of Climate 25: 2129-2145.

Tokinaga, H., S. Xie, A. Timmermann, S. McGregor, T. Ogata, H. Kubota, and Y. Okumura, 2012. Regional Patterns of Tropical Indo-Pacific Climate Change: Evidence of the Walker Circulation Weakening. Journal of Climate 25: 1689-1710.

Vecchi, G.A., B.J. Soden, A.T. Wittenberg, I.M. Held, A. Leetmaa, and H.J. Harrison, 2006. Weakening of the tropical Pacific atmospheric circulation due to anthropogenic forcing. Nature 441: 73-76.

Vecchi, G.A. and B.J. Soden, 2007. Global Warming and the Weakening of the Tropical Circulation. J. Clim. 20: 4316-4340.

Vecchi, G.A., A. Clement, and B.J. Soden, 2008. Examining the Tropical Pacific’s Response to Global Warming. EOS 89(9): 81,83.

Wallace, J.M., 1992. Effect of deep convection on the regulation of tropical sea surface temperature. Nature 357:230-231.

Widlansky, M. J., et al., 2013: Changes in South Pacific rainfall bands in a warming climate. Nature Climate Change 3: 417–423.

Wittenberg, A. T., 2009. Are historical records sufficient to constrain ENSO simulations? Geophysical Research Letters 36: L12702.

Wu, R., and S.P. Xie (2003), On Equatorial Pacific Surface Wind Changes around 1977: NCEP-NCAR Reanalysis versus COADS Observations*. Journal of climate, 16(1), 167-173.

Yeh, S.-W., J.-S. Kug, B. Dewitte, M.-H. Kwon, B.P. Kirtman, and F.-F. Jin, 2009. El Niño in a changing climate. Nature 461: 511-514.

Yeh, S.-W., B. P. Kirtman, J.-S. Kug, W. Park, and M. Latif, 2011. Natural variability of the central Pacific El Nino event on multi-centennial timescales. Geophysical Research Letters 38: L02704.

Yeh, S.-W., Y.-G. Ham, and J.-Y. Lee, 2012. Changes in the tropical Pacific SST Trend from CMIP3 to CMIP5 and its implication of ENSO. Journal of Climate 25: 7764- 7771.

Zhang, M., and H. Song, 2006. Evidence of deceleration of atmospheric vertical overturning circulation over the tropical Pacific. Geophys. Res. Lett. 33:L12701.

APPENDIX A

COMPARISON OF OBSERVED AND SIMULATED TROPICAL CLIMATE TRENDS USING A FORWARD MODEL OF CORAL 18O

Published in Geophysical Research Letters, 38(14), doi: 10.1029/2011GL048224

Reproduced by permission of John Wiley and Sons.

Comparison of observed and simulated tropical climate trends using a forward model of coral δ18O

D.M. Thompson1, T.R. Ault1, M.N. Evans2,1, J.E. Cole1,3, and J. Emile-Geay4

(1) University of Arizona, Department of Geosciences, Gould-Simpson Building #77,

1040 E 4th St. Tucson, AZ 85721

(2) University of Maryland, Dept. of Geology and Earth System Science

Interdisciplinary Center, Geology Bldg, Rm 1120, College Park, MD 20742

(3) Department of Atmospheric Sciences, University of Arizona, Tucson AZ 85721

(4) University of Southern California, Department of Earth Sciences, 3651 Trousdale

Parkway, ZHS 275. Los Angeles, CA 90089 34

Abstract

The response of the tropical Pacific Ocean to future climate change remains highly uncertain, in part because of the disagreement among observations and coupled general circulation models (CGCMs) regarding 20th-century trends. Here we use forward models of climate proxies to compare CGCM simulations and proxy observations to address 20th- century trends and assess remaining uncertainties in both proxies and models. We model coral oxygen isotopic composition (δ18O) in a 23-site Indo-Pacific network as a linear function of sea-surface temperature (SST) and sea-surface salinity (SSS) obtained from historical marine observations (instrumental data) and a multimodel ensemble of 20th- century CGCM output. When driven with instrumental data from 1958 to 1990, the forward modeled corals (pseudocorals) capture the spatial pattern and temporal evolution of the El Niño-Southern Oscillation (ENSO). Comparison of the linear trend observed in corals and instrumental pseudocorals suggests that the trend in corals between 1958 and

1990 results from both warming (60%) and freshening (40%). From 1890 to 1990, the warming/freshening trend in CGCM pseudocorals is weaker than that observed in corals.

Corals display a moderate trend towards a reduced zonal SST gradient and decreased

ENSO-related variance between 1895 and 1985, whereas CGCM pseudocorals display a range of trend patterns and an increase in ENSO-related variance over the same period.

Differences between corals and CGCM pseudocorals may arise from uncertainties in the linear bivariate coral model, uncertainties in the way corals record climate, undersensitivity of CGCMs to radiative forcing during the 20th century, and/or biases in the simulated CGCM SSS fields. 35

1. Introduction

Despite the importance of ENSO to global climate variability, we still lack a clear understanding of how anthropogenic climate change will affect the background state of the tropical Pacific and the amplitude and frequency of ENSO events [e.g., Vecchi et al.

2008]. Instrumental records disagree regarding 20th-century SST trends in the tropical

Pacific [e.g., Vecchi et al. 2008], and CGCM projections of future ENSO behavior differ widely [e.g., Meehl et al., 2007]. Comparison of tropical proxy climate records with

CGCM output over the 20th century provides a way to assess CGCM simulations and constrain predictions for future changes in the background state of the tropical Pacific and the frequency of ENSO events.

Proxy climate records provide an archive of climate variability, but their spatial and temporal coverage is limited, they often reflect a response to a multivariate climate signal, and they may exhibit variance that is unrelated to climate. These proxy limitations add uncertainty to inverse-based methods for reconstructing past climate variability.

Forward modeling of proxy records offers a complementary approach that may curtail the above uncertainties for several reasons. First, it only requires the prediction of the proxy variable. Second, it can be built upon the established dependence of the proxy variable on environmental controls. Finally, it provides a means to directly compare CGCM data and proxy records. For example, previous studies have developed forward models for tree- ring width and isotope composition [e.g., Roden et al., 2000; Vaganov et al., 2006; 36

Tolwinski-Ward et al 2010] and have used these models to study changes in proxy- climate relationships over time [e.g., Anchukaitis et al., 2006].

18 The dependence of coral isotopic composition (δ Ocoral) on environmental conditions is well established and the signal-to-noise ratio in this proxy is high [e.g.,

18 Fairbanks et al., 1997]. Forward modeling of δ Ocoral thus has considerable potential to analyze variations and trends in the tropical Pacific surface ocean climate. For instance,

18 Brown et al. [2006] coupled a linear bivariate model for δ Ocoral, based on SST and

18 18 seawater δ O (δ Osw), with an isotope-enabled CGCM to compare model output with three Indo-Pacific coral records between 1950 and 2000. A subsequent study [Brown et al., 2008] used SST and precipitation from a CGCM to compare ‘pseudocoral’ variability with ENSO variance reconstructed from mid-Holocene corals.

We build on the work of Brown et al. [2006, 2008] by comparing observed

18 th δ Ocoral with that predicted by a linear bivariate model driven by 20 -century instrumental data and CGCM output. Using a network of 23 coral records spanning the

18 Indo-Pacific, we first compare leading patterns of climate variability in observed δ Ocoral

18 18 and δ Ocoral modeled from instrumental data. We refer to these forward-modeled δ Ocoral records as pseudocorals, as in Brown et al. [2008]. We then drive the model with each component separately to diagnose relative importance of salinity and temperature

18 variations in δ Ocoral. We use the model’s ability to simulate the observed spatial and

18 temporal patterns of δ Ocoral to determine whether a more sophisticated treatment of 37

18 th δ Ocoral is necessary. Finally, we compare 20 -century trends in corals and CGCM pseudocorals and assess the uncertainties in this intercomparison.

2. Forward model of coral δ18O

18 Prior experiments and observations have shown that variability in δ Ocoral depends

18 linearly on calcification temperature and local δ Osw at the time of growth [e.g., Epstein et al., 1953], the latter a result of net freshwater flux between the surface ocean and the atmosphere. Furthermore, time series of δ18O variability derived from fast-growing reef corals is offset from isotopic equilibrium with surface conditions [Weber and Woodhead,

18 1972]. We therefore modeled δ Ocoral variability as anomaly time series relative to the

18 average over the full analysis period (1958-1990 or 1890-1990). Because δ Osw measurements are scarce [LeGrande and Schmidt, 2006] and the net freshwater flux

18 affects SSS and δ Osw proportionately via evaporation and condensation processes [Cole

18 and Fairbanks, 1990; Fairbanks et al., 1997], we estimated δ Osw from SSS. This

18 yielded the following model for δ Ocoral anomalies:

푝 푒𝑢푑표푐표𝑟𝑎푙 = 푎 푆푆푇 + 푎2푆푆푆 (1),

18 where a2 was specified through basin-scale δ Osw vs. SSS regression estimates (‰/PSU,

Table A-S1; [LeGrande and Schmidt, 2006]). We specified a1 based on the experimental and theoretical dependence of oxygen isotopic equilibrium on the temperature of carbonate formation [e.g., Epstein et al., 1953], which has been observed in well-studied coral genera [e.g., Evans et al., 2000; Juillet-Leclerc and Schmidt, 2001; Lough et al.,

2004]. Although the slope of this relationship may range from -0.10 to -0.34 ‰/ºC at 38

individual sites [Evans et al., 2000], studies that synthesized multiple locations report slopes of -0.2 (10 sites, [Evans et al., 2000]) and -0.22 ‰/ºC (19 sites, [Lough et al.,

2004]), close to the inorganic slope of -0.22‰/ºC. We selected a slope of a1= -0.22 (+/-

18 0.02) ‰/ºC for the SST- δ Ocoral relationship.

3. Input datasets

We tested the forward model using SST and SSS from instrumental data products

(SST: ERSSTv2, [Smith and Reynolds, 2004], ERSSTv3 [Smith et al., 2008], Kaplan ext.v2 [Kaplan et al., 1998], HadISST [Rayner et al., 2003]; SSS: SODA [Carton and

Giese, 2008], Carton GOA Beta 7 [Carton et al., 2000]). We then ran the forward model with SST and SSS from 20th-century simulations of several AR4 CGCMs, chosen to reflect the range of 20th-century variability, trends, and ENSO skill exhibited by the full suite of AR4 models [e.g., Meehl et al., 2007; Guilyardi et al., 2009]. We used eq. 1 and observed or simulated SST and SSS from the nearest gridbox to model monthly δ18O for each coral site. The latitude/longitude grid resolution ranged from 0.5 x 0.5 (e.g., SODA) to 5 x 5 (e.g., Kaplan ext.v2); when necessary, we regridded data to the same resolution using box averaging. Each monthly pseudocoral series was annually averaged for further

18 analysis. To assess the relative contribution of SST and SSS in δ Ocoral, pseudocorals were also modeled using SST and SSS separately (see Table A-S2 for all combinations tested).

4. Comparison of observed and pseudocoral networks

18 The observational target was a dataset of 23 annual δ Ocoral anomaly records, spanning ≥ 90% of the 1850-1990 interval [Ault et al., 2009, Table A-S1]. Unlike Ault et 39

al. [2009], we did not detrend the δ18O series because the long-term trends were of prime interest. We used singular value decomposition (SVD) of the δ18O covariance matrix to determine the spatiotemporal patterns of variability within observed and pseudocoral

δ18O networks over the common time interval, 1958-1990 for the instrumental period and

1890-1990 for the CGCMs. The “rule N” test [Overland and Preisendorfer, 1982] was used to identify significant eigenvectors above the “white noise floor” at the 95% confidence level, where the white noise floor represents the expected eigenvalues if there were no spatial structure.

Two significant eigenvectors emerged from the observed coral network, explaining

~30 and ~17% of the variance, respectively, over the 1958-1990 period. We interpreted the pattern of these eigenvectors as the signature of ENSO and a secular trend in the observational network, respectively (Figure A-S1). The secular trend emerged as the first eigenvector over the 1890-1990 period, explaining 28% of the variance, while ENSO was the second eigenvector and explained 17% of the variance. The ENSO and trend patterns were also significant and explained the leading fractions of variance in all pseudocoral networks (Table A-S2). Varimax rotation of the significant principal components (PCs) had no marked effect on ENSO or the trend.

We examined whether instrumental and CGCM pseudocorals captured the leading modes of variability in the observed coral network. To do this, we compared the spatial and temporal patterns of ENSO and the trend between networks over contemporaneous time intervals (see Supplementary Methods 1.1). 40

We then compared observed coral and CGCM pseudocoral trends in ENSO- related variance and tropical Pacific “mean state” over the 20th century [after Meehl et al.,

2007] (Figure A-2, see Supplementary Methods 1.2). To calculate the mean state trend

(Figure A-2, x-axis), we correlated the spatial expression of ENSO with the spatial pattern of the 20th-century trend. We referred to the trend as “El Niño-like” if the spatial pattern of the trend correlated positively with the pattern of warm-phase ENSO [after

Meehl et al., 2007]. We then calculated the change in ENSO variability (y-axis) as the ratio of ENSO-related interannual variability between the last half and first half of the analysis period, such that a ratio greater than one indicates an increase in ENSO-related variability. We performed sensitivity analyses for the mean state and ENSO-related variance calculations using randomly selected subsets of sites (for mean state error) and years (for variance error) from the original data set.

5. Results

5.1 Pseudocorals derived from instrumental observations

With one exception (Carton GOA SSS pseudocorals), the ENSO PC from instrumental pseudocorals was significantly related to the ENSO PC from the observed coral network over the 1958-1990 period (Table A-S2, Figure A-S1b). Comparisons of the ENSO PC correlation fields illustrate that instrumental pseudocorals also captured the spatial pattern of ENSO variability over this period (Table A-S3). We found that for pseudocorals generated with both SST and SSS, the ENSO correlation field differed significantly from the observed coral field in only three (out of ten) cases: HadISST & 41

SODA vs. SST and SSS, and Kaplan ext.v2-SST & SODA vs. SSS (Table A-S3).

Instrumental pseudocorals also captured the observed trend between 1958 and 1990

(Table A-S2). The relationship between the observed and pseudocoral trend PCs was strongest for the pseudocoral networks that were modeled with both SST and SSS.

Significant relationships between the observed and pseudocoral trend PCs existed despite differing superimposed interannual variability (i.e., the time evolution of the trends), indicating strong consistency of the underlying linear trends. After isolating the linear trend with a linear regression, the residuals were normally distributed around a mean of zero for all instrumental pseudocoral networks. The observed coral trend residuals also displayed a zero mean (Z= 0.477, df=32, P=0.684), despite being positively skewed by the superimposed interannual anomalies. These results indicate that the linear trends, estimated from regression through the trend PCs, were a good fit to the underlying secular trend.

We focused our analysis on the linear trend component as an approximation of the secular trend. This trend toward more negative δ18O values was significant in all coral networks. Additionally, the RMSE between regression fields of observed and pseudocoral linear trends regressed on SST and SSS were indistinguishable from zero. In both observed corals and pseudocorals, the regression fields suggested that this linear trend reflects increasing SST throughout most of the tropical Indo-Pacific. SSS played an important role in the trend regionally, particularly in the western and southwestern

Pacific. Thus, the addition of SSS improved the ability of the model to capture the full magnitude of the observed coral trend between 1958 and 1990 (-0.167 ‰/decade, Figure 42

A-1a). This improvement cannot be explained simply by the addition of another component to the model, as the mean trend of 100 pseudocoral networks modeled with

SST and noise with the same variance and lag 1 autocorrelation as SSS was similar to that observed in the SST only pseudocoral networks (-0.105 ± 0.018 ‰/decade).

Although the magnitude of the coral trend was still about 20% larger than observed in any SST and SSS pseudocoral networks (max of -0.138‰/decade, Figure A-1a, S1f), the difference between the observed and pseudocoral trend slopes was not significant. The mean difference between the observed and pseudocoral trend at the 23 coral sites was also not significantly different from zero, suggesting that the linear bivariate model also captured the trends observed at the 23 coral sites. Based on the relative amplitude of the trend in SSS-only and SST-only pseudocorals and that observed over the 1958-1990 period, approximately 40% of the coral trend (-0.045 to -0.072‰/decade) was explained by salinity and approximately 60% of the trend (-0.098 to -0.11 ‰/decade) was explained by temperature.

5.2. Pseudocorals derived from CGCM output

Several studies have established that AR4 CGCMs generally simulate the spatial pattern of ENSO, albeit excessively locked to a two-year periodicity [e.g., Guilyardi et al., 2009]. We focused here on trends in the tropical Pacific mean state and in ENSO- related variability because these are harder to constrain using instrumental records alone.

As observed in corals, all CGCM pseudocoral networks contained a significant linear trend toward more negative δ18O values between 1890 and 1990. The linear trend residuals were normally distributed with a mean of zero, indicating again that the linear 43

trend was a good approximation of the underlying secular trend. However, the slope of the linear trend (‰/decade) in all CGCM pseudocoral networks was lower than observed in corals between 1890 and 1990 (Figure A-1b). Site by site comparison suggests that this discrepancy may result from weaker trends in the central western Pacific in the

CGCM pseudocorals than observed, particularly at Maiana and Jarvis Islands.

We also found that corals and CGCM pseudocorals disagreed regarding 20th- century trends in tropical Pacific mean state and ENSO-related variance (Figure A-2).

Corals displayed a reduction in ENSO-related variance and a moderate El Niño-like mean state trend. In contrast, CGCM pseudocorals displayed a range of mean state trends and suggested an increase in variance over this same period. Although HadCM3 pseudocorals displayed a trend pattern that was most similar to that observed, HadCM3 pseudocorals also showed a much larger reduction in ENSO-related variance than observed, and the trend in HadCM3 pseudocorals over this period was weak (-

0.043‰/decade, Figure A-1) and sensitive to the selection of sites in the underlying network. When the network included mainly ENSO-sensitive sites, HadCM3 displayed a stronger La Niña-like trend pattern (as in projections of future change [Meehl et al.

2007]).

6. Discussion

Our model captured the spatial and temporal pattern of ENSO and the linear trend

18 observed in corals from 1958 to 1990 and demonstrated that the δ Ocoral trend results from both warming and freshening. These results are consistent with observational studies showing freshening in the tropical Pacific [e.g., Cravatte et al., 2009; Nurhati et 44

al., 2009]. None of the CGCM pseudocoral networks captured the magnitude of the trend, the change in mean state, or the change in ENSO-related variance observed in the coral network over the 1890-1990 period. The negative δ18O trend was weaker in CGCM pseudocorals than observed, and whereas corals displayed a moderately El Niño-like trend in the Pacific, there was little agreement among CGCMs regarding the spatial pattern of the trend. Finally, the corals displayed a reduction in ENSO-related variance between the first and last half of the analysis period, whereas CGCM pseudocorals generally displayed an increase.

The discrepancies between observed and CGCM pseudocoral trends may stem from uncertainty in the observational coral records, the CGCM output (SST and SSS), and/or the way we translate CGCM output into coral records. Results from sensitivity analyses of the trend pattern suggest that network biases stemming from the site locations and/or time period analyzed are unlikely to have caused this discrepancy, as none of the randomly selected subsets from the CGCM pseudocoral networks overlap with any of the subsets from the observed coral network. Also, the effect of subsampling on the resulting patterns was generally small (see error bars in Figure A-2). Finally, it is unlikely that the bivariate linear approximation of the proxy system was the primary cause of the discrepancy, as instrumental pseudocorals were able to capture the interannual variability and trend observed in corals between 1958 and 1990.

Although not significant, the magnitude of the trend was lower in the instrumental pseudocorals than in real corals. This discrepancy appears strongest at sites where instrumental observations are limited (e.g., at the Maiana site), suggesting that these 45

gridded instrumental data products may underestimate SSS and/or SST trends. On the other hand, the discrepancy between observed and pseudocoral trends could be caused by biologically-mediated kinetic isotope effects that may occur in corals under unusually stressful conditions. Although earlier work demonstrated isotopic excursions with extremely slow growth rates [McConnaughey, 1989], we have found no relationship between extension rate and δ18O over the range of growth observed in the corals studied here.

To determine whether the observed-modeled trend discrepancy could be attributed to

SST or SSS, the trend-ENSO pattern and ENSO variance analysis was repeated with

GCCM temperature and salinity separately at each of the 23 coral sites (Figure A-2, colored symbols). The results suggest that SSS played an important role in the simulated trend pattern. The discrepancy between coral and CGCM pseudocoral trends may therefore be driven by differences in the salinity recorded by corals and that simulated by

CGCMs. A salinity bias in the coral records, due perhaps to local effects or the way SSS is recorded in corals, is not likely the source of the discrepancy because: (1) the

18 agreement of coral records on a regional scale suggests that local SSS biases on δ Ocoral are minor, (2) the addition of SSS improves the agreement between instrumental pseudocorals and observations, and (3) instrumental pseudocorals and observed corals agree closely.

On the other hand, CGCM SSS often displayed different mean state and ENSO- related variance trends than observed for temperature (Figure A-2). Further analyses 46

revealed that the discrepancy between observed corals and CGCM pseudocorals may be caused by biases in the CGCM representation of the hydrological cycle. First, while instrumental SST and SSS displayed a negative relationship throughout most of the tropical Indo-Pacific (Figure A-3a), as expected from temperature-driven tropical convection, all but one CGCM analyzed here displayed positive SST-SSS correlations in the eastern and central equatorial Pacific (Figure A-S2). The average CGCM correlation pattern, weighted by the model’s ability to capture tropical climate variability [Gleckler et al., 2008], displayed a region of significant positive SST-SSS correlations in the central to eastern tropical Pacific (Figure A-3b). A similar pattern was observed in the unweighted composite. Second, the magnitude and direction of SSS trends simulated by these CGCMs were very different from instrumental SSS trends and from each other

(cancelling each other in many regions of the composite) (Figure A-3c,d, Figure A-S2).

Thus, biases in the SSS trend and variability at the 23 coral sites (Figure A-3d) may contribute to the discrepancy between observed and CGCM simulated coral trends over the 20th century.

7. Conclusions and future work

18 A linear temperature- and salinity-driven model for δ Ocoral captured the spatial and temporal pattern of ENSO and the linear trend observed in corals between 1958 and

1990. The negative δ18O trend observed in corals had substantial contributions from both general warming (~60%) and regional freshening (~40%). When we drove this proxy model with CGCM output, we found differences between observed and simulated coral 47

trends over the 20th century. Our work highlights potential biases in CGCM-simulated

SSS fields, suggesting that the response of the tropical hydrological cycle to 20th-century climate forcing needs to be further investigated to improve our understanding of the evolution of tropical Pacific climate. Discrepancies between coral and CGCM trends may also result from variance in the observed coral dataset that is unexplained by the

18 linear bivariate model. Further investigation of potential nonlinearites in how δ Ocoral tracks climate, and coupling the proxy model presented here with isotope-enabled climate models, will provide useful tools to explore these discrepancies.

Acknowledgements

We thank Joellen L. Russell, Jonathan T. Overpeck, Alexander W. Tudhope, J. Warren

Beck, David Noone, and Heidi Barnett for helpful comments and support on this project.

This research was supported by NOAA’s Climate Change Data and Detection Program

(award NA10OAR4310115 to JEG, MNE, and DMT and award NA08OAR4310682 to

JEC), by the University of Arizona Department of Geosciences, and by the Institute of the Environment at the University of Arizona.

Reference list

Anchukaitis, K. J., et al. (2006), Forward modeling of regional scale tree-ring patterns in

the southeastern United States and the recent influence of summer drought, Geophys.

Res. Lett., 33, L04705, doi:10.1029/2005GL025050. 48

Ault, T.R., et al. (2009), Intensified decadal variability in tropical climate during the late

19th century, Geophys. Res. Lett., 36, L08602, doi:10.1029/2008GL036924.

Brown, J., I. Simmonds, and D. Noone (2006), Modeling δ18O in tropical precipitation

and the surface ocean for present-day climate, J. Geophys. Res., 111, D05105,

doi:10.1029/2004JD005611.

Brown, J., et al. (2008), Mid-Holocene ENSO: Issues in quantitative model-proxy data

comparisons, Paleoceanography, 23, PA3202, doi:10.1029/2007PA001512.

Carton, J.A., and B.S. Giese (2008), A reanalysis of ocean climate using Simple Ocean

Data Assimilation (SODA), Mon. Weather Rev., 136, 2999–3017,

doi:10.1175/2007MWR1978.1.

Carton, J.A., et al. (2000), A Simple Ocean Data Assimilation analysis of the global

upper ocean 1950-1995, Part 1: methodology, J. Phys. Oceanogr., 30, 294-309.

Cole, J. E., and R. G. Fairbanks (1990), The Southern Oscillation recorded in the δ18O of

corals from Tarawa Atoll, Paleoceanography, 5(5), 669–683,

doi:10.1029/PA005i005p00669.

Cravatte, S., et al. (2009), Observed freshening and warming of the western Pacific

Warm Pool, Clim. Dyn., 33(4), 565-589, doi:10.1007/s00382-009-0526-7.

Epstein, S., et al. (1953), Revised carbonate-water isotopic temperature scale, Geol. Soc.

America Bull., 64(11), 1315-1326. 49

Evans, M. N., A. Kaplan, and M.A. Cane (2000), Intercomparison of coral oxygen

isotope data and historical sea surface temperature (SST): Potential for coral-based

SST field reconstructions, Paleoceanography, 15(5), 551-563,

doi:10.1029/2000PA000498.

Fairbanks, R.G., et al. (1997), Evaluating climate indices and their geochemical proxies

measured in corals, Coral Reefs, 16(5), S93-S100, doi:10.1007/s003380050245.

Gleckler, P. J., et al. (2008), Performance metrics for climate models, J. Geophys. Res.,

113, D06104, doi:10.1029/2007JD008972.

Guilyardi, E., et al. (2009), Understanding El Niño in ocean-atmosphere general

circulation models: progress and challenges, Bull. Amer. Meteorol. Soc., 90(3), 325-

339.

Juillet-Leclerc, A., and G. Schmidt (2001), A calibration of the oxygen isotope

paleothermometer of coral aragonite from Porites, Geophys. Res. Lett., 28, 4135–

4138.

Kaplan, A., et al. (1998), Analyses of global sea surface temperature 1856–1991, J.

Geophys. Res., 103(C9), 18,567–18,589, doi:10.1029/97JC01736.

LeGrande, A. N., and G. A. Schmidt (2006), Global gridded data set of the oxygen

isotopic composition in seawater, Geophys. Res. Lett., 33, L12604,

doi:10.1029/2006GL026011. 50

Lough, J. M. (2004), A strategy to improve the contribution of coral data to high-

resolution paleoclimatology, Palaeogeogr., Palaeoclimatol., Palaeoecol., 204(1-2),

115-143, doi:10.1016/S0031-0182(03)00727-2.

McConnaughey, T. (1989), 13C and 18O isotopic disequilibrium in biological carbonates:

I. Patterns, Geochim. Cosmochim. Acta, 53, 151-162, doi:10.1016/0016-

7037(89)90282-2.

Meehl, G.A., et al. (2007), Global climate projections, in Climate Change 2007: the

physical science basis. Contribution of Working Group I to the Fourth Assessment

Report of the Intergovernmental Panel on Climate Change, edited by S. Solomon, et

al., pp. 747–845, Cambridge University Press, Cambridge.

Nurhati, I. S., et al. (2009), Late 20th century warming and freshening in the central

tropical Pacific, Geophys. Res. Lett., 36, L21606, doi:10.1029/2009GL040270.

Overland, J.E., and R.W. Preisendorfer (1982) A significance test for principal

components applied to a cyclone climatology, Mon. Weather Rev., 110(1), 1-4.

Rayner, N. A., et al. (2003), Global analyses of sea surface temperature, sea ice, and

night marine air temperature since the late nineteenth century, J. Geophys. Res.,

108(D14), 4407, doi:10.1029/2002JD002670.

Roden, J.S., G. Lin, and J.R. Ehleringer (2000) A mechanistic model for interpretation of

hydrogen and oxygen isotope ratios in tree-ring cellulose, Geochim. Cosmochim.

Acta, 64(1), 21-35, doi:10.1016/S0016-7037(99)00195-7. 51

Smith, T.M. and R.W. Reynolds (2004) Improved extended reconstruction of SST (1854–

1997), J. Climate, 17, 2466–2477.

Smith, T.M., et al. (2008) Improvements to NOAA's historical merged land-ocean

surface temperature analysis (1880-2006), J. Climate, 21, 2283-2296.

Tolwinski-Ward, S., et al. (2010), An efficient forward model of the climate controls on

interannual variation in tree-ring width, Clim. Dyn., 1-21, doi:10.1007/s00382-010-

0945-5

Vaganov, E.A., M.K. Hughes, and A.V. Shashkin (2006), Growth Dynamics of Tree

Rings: Images of Past and Future Environments, Springer, New York.

Vecchi, G. A., A. Clement, B. J. Soden, (2008), Examining the tropical Pacific's response

to global warming, Eos Trans. AGU, 89(9), 81, doi:10.1029/2008EO090002.

Weber, J. N., and P. M. J. Woodhead (1972), Temperature Dependence of Oxygen-18

Concentration in Reef Coral Carbonates, J. Geophys. Res., 77(3), 463–473,

doi:10.1029/JC077i003p00463. 52

Figure A-1. (a) Magnitude of the trend slope (‰/decade, computed from a simple linear regression through the trend PC) in corals (far left) and in instrumental pseudocorals over the 1958-1990 period. Error bars denote ± 1 standard deviation. (b) As in (a) for the trend in corals and CGCM pseudocorals over the 1890-1990 period.

Figure A-2. Change in mean state (x-axis) and ENSO-related variability (y-axis) [after Meehl et al., 2007] within the Pacific (120ºE to 80ºW) between 1895 and 1985 in corals (black asterisk) and CGCM (SST & SSS) pseudocorals (black symbols). The symbols mark the mean and error bars denote the 95% confidence interval of 1000 independently sub-sampled variance ratios and 50 independently sub-sampled pattern correlations. Colored symbols represent analysis of CGCM SST (blue) and SSS (red) at the 23 coral sites. The full CGCM SST field was also analyzed, and the results were similar to that observed at the 23 coral sites. (Note: for CGCMs with no SSS symbol, SSS did not display a significant trend component) 54

Figure A-3. Left side: Correlation of SST and SSS from 1958 to 1990 over the tropical Indo-Pacific in (a) instrumental SST and SSS: ERSSTv2 and SODA and (b) the 7 CGCMs analyzed here (weighted average). Right side: SSS trend (PSU/decade) from 18 1958 to 1990 in (c) SODA SSS and (d) the 7 CGCMs (weighted average). The δ Ocoral trends (‰/decade) over this period were also plotted for comparison. Hatching indicates significant correlation (a), trend values (c), or gridboxes for which the mean value in the 7 models was significantly different from zero (b,d) at the 95% confidence level. The models were interpolated to a common 2x2 degree latitude/longitude grid resolution and weighted based on their ability to capture tropical climate variability [1-Model Variability Index (MVI); Gleckler et al., 2008] prior to averaging and significance testing. (see Figure A-S2 for individual model patterns).

S1. Supplementary methods:

S1.1. Forward modeled ‘pseudocoral’- observed coral comparisons

To assess the ability of each pseudocoral network to capture the patterns of variability in the observed coral network, we compared the spatial and temporal pattern of ENSO and the trend. First, we obtained the spatial expression of the significant modes of δ18O variability by correlating the PCs with the SST and SSS fields. We then regressed the linear trend onto the fields to determine the magnitude of the trend (in

°C/year and PSU/year) throughout the Indo-Pacific. We calculated the root mean square error (RMSE) between the observed and pseudocoral correlation and regression fields as:

N 2 , (A1), ()coeff / ()()obs i coeff pseudo i N i1 where coeff represents the correlation coefficient (r) or regression coefficient (β) over N grid boxes indexed by i. In each case, we projected the ENSO PC and linear trend onto the corresponding SST and SSS fields, with ERSSTv2-SST and SODA-SSS used as default fields for the other variable in SST or SSS only pseudocoral networks. Second, to determine whether the psuedocorals captured the temporal evolution of ENSO and the trend observed in the corals, we regressed the pseudocoral PCs on the observed PCs. If the pseudocorals captured the evolution of ENSO and the trend through time, a significant relationship with a slope near unity would be expected. Finally, we used an analysis of covariance [Zar, 1999] to compare the slope of the shared linear trend in the observed and pseudocoral networks (‰ per decade). To address whether differences in 56

the magnitude of the trend could be attributed to discrepancies in certain regions of the tropical Indo-Pacific, we also compared the magnitude of the trend slope at each coral site. A simple t-test was used to determine if the mean difference between the observed and pseudocoral trend at the 23 sites was significantly different from zero.

S1.2. Mean state and ENSO-related variance

We compare 20th-century trends in mean state and ENSO-related variance, following an approach similar to that of Meehl et al. [2007] [after van Oldenborgh et al.,

2005; Yamaguchi and Noda, 2006]. First, we high-pass filtered the ENSO PC by convoluting an 11-year normalized hamming window with the ENSO PC to isolate the interannual variance. Because this filtering process removes 5 years from either end of the record, the following analyses were performed between 1895 and 1985. We determined the spatial expression of ENSO and the trend at the coral sites by regressing the high-pass filtered ENSO PC and linear trend onto the observed or pseudocoral records at the 23 coral sites. We calculated the mean state change in the tropical Pacific

(x-axis) as the correlation between the ENSO and linear trend regression coefficients at the 14 Pacific coral sites (80ºW-120ºE). A positive correlation coefficient indicates spatial similarity between the trend and the ENSO pattern and is thus referred to as an “El

Niño-like” trend [after Meehl et al., 2007]. Note that this terminology is used only to refer to the spatial pattern of the trend, with no inferences for the underlying dynamics.

Finally, we calculated the change in ENSO variability (y-axis) as the ratio of the standard deviation of the high-pass filtered ENSO PC in the last 45 years (1941-1985) to the 57

standard deviation of the filtered ENSO PC in the first 45 years (1895-1939) of the analysis. A ratio of 1 indicates that there was no change in the standard deviation of the

ENSO PC (ENSO-related variance) between the first and last 45 years of the analysis.

To assess the sensitivity of the ENSO-related variance and mean state trends to the underlying network and time period covered by the analysis, we performed sensitivity analyses by randomly selecting subsets from the original data set. First, we randomly selected thirty years from the first and last 45 years of the high-pass filtered ENSO PC and calculated the standard deviation ratio (y-axis) for each subset. We then assessed the sensitivity of the mean state trend to the underlying network of coral sites by randomly selecting twenty sites from the full 23-site network for SVD analysis. For each subset, we selected the ENSO and trend PCs from the consecutive leading significant principal components (PCs 1-3) and the mean state trend was calculated as above. We defined the trend as the PC that displayed the greatest slope (β) when regressed on time, and ENSO as the remaining PC that displayed the strongest relationship with (instrumental or modeled) Nino 3.4 SSTs. We excluded the subset from the mean state calculation if

ENSO and the trend were not separated into independent PCs (observed in a few GFDL cm2.1 subsets). We performed further sensitivity tests by varying the number of subsets analyzed (between 10 and 1000), by performing the variance calculation on the full 100- year period (before filtering the ENSO PC), by varying hamming window length, and by changing the criteria for selection of the PCs (e.g., the number considered, statistics for selection, and whether significance was required). Varying these input criteria did not influence the interpretation of the results. Therefore, we selected the most conservative 58

approach, where significance of the PCs was required, the ENSO PC was filtered with an

11-year normalized hamming window to maintain only variability up to 10 years, and the largest number of subsets that maintained independence of the samples was used to quantify sensitivity (1000 subsets of time and 50 subsets of sites).

18 Table A-S1: Slope of the δ Osw-salinity relationship (‰/PSU, from LeGrande and Schmidt 2006) for the regions containing coral records used in this analysis [Ault et al., 2009].

Region Latitudes Longitudes Slope (+/- σ) Tropical Pacific 5S to 13N 70W to 120E 0.27 (+/- 0.006) Chiriqui, Panama 7ºN 82ºW Linsley et al., 1994 Urvina Bay, Galapagos 0ºS 91ºW Dunbar et al., 1994 Jarvis Island 0ºS 160ºW Tudhope et al., in prep. Maiana Atoll, Kiribati 0ºN 173ºE Urban et al., 2000 Double Reef, Guam 13ºN 144ºE Asami et al., 2005 Bunaken Island, Indonesia 1ºN 124ºE Charles et al., 2003 South Pacific 5S to 28S 70W to 113E 0.45 (+/- 0.028) Moorea, French Polynesia 17ºS 149ºW Boiseau et al., 1998 Rarotonga, Cook Islands 21ºS 159ºW Linsley et al., 2000 Palmerston, Cook Islands 18ºS 163ºW Tudhope et al., in prep. Savusavu Bay, Fiji 17ºS 178ºE Bagnato et al., 2005 Espiritu Santo, Vanuato 15ºS 167ºE Quinn et al., 1996 New Caledonia 22ºS 166ºE Quinn et al., 1998 Abraham Reef, Australia 22ºS 153ºE Druffel and Griffin, 1999 Bramble Cay, Papua New Guinea 9ºS 144ºE Cole, in preparation Houtman Abrolhos, Australia 28ºS 113ºE Kuhnert et al., 1999 Indian Ocean 23S to 20N 120E to 38E 0.16 (+/- 0.004) Bali, Indonesia 8ºS 115ºE Charles et al., 2003 Mahe, Seychelles 4ºS 55ºE Charles et al., 1997 Mentawai, Sumatra 2ºS 99ºE Abram et al., 2008 La Reunión 21ºS 55ºE Pfeiffer et al., 2004 Ifaty, Madagascar 23ºS 43ºE Zinke et al., 2004 Malindi, Kenya 3ºS 40ºE Cole et al., 2000 Barnett, 2006; Barnett et Tutia, Tanzania 8ºS 39ºE al., in prep. Barnett, 2006; Barnett et Zanzibar, Tanzania 6ºS 39ºE al., in prep.

Table A-S2: Statistics for the regression of the observed coral ENSO PC (PC2) and the observed coral trend PC (PC1) with the pseudocoral ENSO and trend PCs, where bold indicates significance at the 95% confidence level and bold italics represent significance at the 99% confidence level. Pseudocoral network % variance R2 F P slope input datasets explained Tren ENS ENS SST SSS ENSO Trend ENSO Trend ENSO Trend Trend d O O ERSSTv2 SODA 51.4% 14.9% 0.497 0.445 30.7 24.8 <0.001 <0.001 0.667 0.862 ERSSTv2 CartonGOA 53.6% 16.6% 0.488 0.317 29.5 14.4 <0.001 <0.001 0.701 0.751 ERSSTv3 SODA 51.7% 14.2% 0.487 0.444 29.5 24.7 <0.001 <0.001 0.651 0.871 HadISST SODA 47.2% 14.4% 0.427 0.365 23.1 17.8 <0.001 <0.001 0.605 0.743 Kaplan SODA 44.1% 14.3% 0.457 0.650 26.1 57.6 <0.001 <0.001 0.695 1.07 ext.v2 ERSSTv2 52.7% 17.9% 0.474 0.333 27.9 15.5 <0.001 <0.001 0.788 0.833 ERSSTv3 52.9% 16.9% 0.464 0.272 26.7 11.6 <0.001 0.0018 0.766 0.763 HadISST 47.6% 17.3% 0.375 0.344 18.6 16.3 <0.001 <0.001 0.688 0.802 Kaplan 42.0% 18.7% 0.438 0.400 24.1 20.7 <0.001 <0.001 0.824 0.869 ext.v2 SODA 40.1% 40.1% 0.433 0.148 23.7 5.37 <0.001 0.0272 1.80 0.772 CartonGOA 20.1% 44.2% 0.0435 0.115 1.41 4.02 0.244 0.0538 1.61 1.29

Table A-S3: Root mean square error (RMSE) between the observed and pseudocoral ENSO correlation fields, calculated as:

N 2 ()coeff / (obs )( ipseudo coeffN ) i i1 , where coeff represents the correlation coefficient (r) and N is the number of grid boxes (i). In each case, the ENSO PC was projected onto the SST and SSS fields corresponding to the input for the pseudocoral network, with ERSSTv2-SST and SODA-SSS used as default fields for the other variable in SST or SSS only pseudocoral networks. Values in bold indicate RMSEs which were less than the 950th lowest RMSE of 1000 random time series correlated with the underlying SST or SSS field (i.e., the two fields were not significantly different at the 95% CI). Pseudocoral network ENSO PC- ENSO PC- input datasets SST RMSE SSS RMSE SST SSS ERSSTv2 SODA 0.211 0.224 ERSSTv2 CartonGOA 0.202 0.222 ERSSTv3 SODA 0.221 0.231 HadISST & SODA 0.232 0.260 Kaplan ext.v2 SODA 0.232 0.246 ERSSTv2 0.202 0.231 ERSSTv3 0.215 0.238 HadISST 0.316 0.382 Kaplan ext.v2 0.232 0.253 SODA 0.237 0.180 CartonGOA 0.302 0.415

Figure A-S1. (a) Correlation of the ENSO PC observed in corals over the 1958-1990 18 period (b, black) with the δ Ocoral record from each coral site. (c) Same as in (a) but for the high-pass filtered ENSO PC from corals over the 1890-1990 period (d). (e) 18 Regression coefficient from the regression of the δ Ocoral records with the linear trend over the 1958-1990 period, estimated from linear regression of the trend PC (f). (g) Same as in (e) but for the linear trend over the 1890-1990 period (h). The ENSO PC (b) and linear trend (f) from pseudocorals modeled with both SST and SSS are shown in gray for comparison.

Figure A-S2. (left) Correlation of SST and SSS and (right) the SSS trend (PSU per decade) between 1958 and 1990 over the tropical Indo-Pacific in (a) Observational SST and SSS: ERSSTv2 and SODA, (b) GFDL cm 2.0, (c) GFDL cm 2.1, (d) NCAR CCSM3, (e) NCAR PCM1, (f) GISS eh, (g) GISS er, and (h) HadCM3. Only significant correlations and trends are shown.

Supplemental references:

Abram, et al. (2008) Recent intensification of tropical climate variability in the Indian Ocean, Nature Geosciences, doi:10.1038/ngeo357.

Asami, R., et al. (2005), Interannual and decadal variability of the western Pacific sea surface condition for the years 1787-2000: Reconstruction based on stable isotope record from a Guam coral, J. Geophys. Res. Oceans, 110(C5), doi:10.1029/2004JC002555.

Bagnato, S., et al. (2005), Coral oxygen isotope records of interdecadal climate variations in the South Pacific Convergence Zone region, Geochemistry Geophysics Geosystems, 6, doi:10.1029/2004GC000879.

Barnett H.B., et al., Unpublished data.

Barnett H.B., (2006), 20th Century climate variability in the tropical Indian ocean from a new network of coral oxygen isotope chronologies, M.S., Dept. of Geosciences, University of Arizona.

Boiseau, M., et al. (1998), Atmospheric and oceanic evidences of El Niño Southern scillation events in the south central Pacific Ocean from coral stable isotopic records over the last 137 years, Paleoceanography, 13(6), 671-685.

Charles, C. D., et al. (1997), Interaction between the ENSO and the Asian monsoon in a coral record of tropical climate, Science, 277(5328), 925-928.

Charles, C.D., et al. (2003), Monsoon-tropical ocean interaction in a network of coral records spanning the 20th century, Marine Geology, 201, 207-222.

Cole, J. E., et al. (2000), Tropical Pacific forcing of decadal SST variability in the western Indian Ocean over the past two centuries, Science, 287(5453), 617-619.

Cole, J.E., et al., Unpublished data.

Druffel, E. R. M., and S. Griffin (1999), Variability of surface ocean radiocarbon and stable isotopes in the southwestern Pacific, J. Geophys. Res. Oceans, 104(C10), 23607-23613.

Dunbar, R.B., G.M.Wellington, M.W. Colgan, and P.W. Glynn (1994), Eastern Pacific Sea Surface Temperature since 1600 A.D.: The d18O Record of Climate Variability in Galapagos Corals, Paleoceanography, 9, 291-315.

Kuhnert, H., et al. (1999), A 200-year coral stable oxygen isotope record from a high- latitude reef off western Australia, Coral Reefs, 18(1), 1-12.

Linsley, B. K., et al. (1994), A coral-based reconstruction of intertropical convergence zone variability over Central America since 1707, J. Geophys. Res. Oceans, 99(C5), 9977-9994.

Linsley, B. K., et al. (2000), Decadal sea surface temperature variability in the subtropical South Pacific from 1726 to 1997 AD, Science, 290(5494), 1145-1148.

Pfeiffer, M., et al. (2004), Oceanic forcing of interannual and multidecadal climate variability in the southwestern Indian Ocean: Evidence from a 160 year coral isotopic record (La Reunion, 55 degrees E, 21 degrees S), Paleoceanography, 19(4), doi:10.1029/2003PA000964.

Quinn, T. M., et al. (1996), New stable isotope results from a 173-year coral from Espiritu Santo, Vanuatu, Geophys. Res. Lett., 23(23), 3413-3416.

Quinn, T. M., et al. (1998), A multicentury stable isotope record from a New Caledonia coral: Interannual and decadal sea surface temperature variability in the southwest Pacific since 1657 AD, Paleoceanography, 13(4), 412-426.

Tudhope, A.W., et al., Unpublished data.

Urban, F. E., J. E. Cole, and J. T. Overpeck (2000), Influence of mean climate change on climate variability from a 155-year tropical Pacific coral record, Nature, 407(6807), 989–993.

van Oldenborgh, G.J., S.Y. Philip, and M. Collins (2005), El Niño in a changing climate: a multi-model study, Ocean Sci., 1, 81–95.

Yamaguchi, K., and A. Noda (2006), Global warming patterns over the North Pacific: ENSO vs. AO, J. Meteorol. Soc. Jpn., 84(1), 221– 241, doi:10.2151/jmsj.84.221.

Zar, J.H. (1999), Comparing Simple Linear Regression Equations, in Biostatistical analysis, 4th ed., pp. 360-376, Prentice Hall, Upper Saddle River, New Jersey. 66

Zinke, J., et al. (2004), ENSO and Indian Ocean subtropical dipole variability is recorded in a coral record off southwest Madagascar for the period 1659 to 1995, Earth Planet. Sci. Lett., 228(1-2), 177-194.

APPENDIX B

ENHANCED E-W TEMPERATURE GRADIENT ACROSS THE DATE LINE OVER THE 20th CENTURY INFERRED FROM CENTRAL PACIFIC CORAL RECORDS

To be submitted to the professional journal: Paleoceanography

ENHANCED E-W TEMPERATURE GRADIENT ACROSS THE DATE LINE OVER THE 20th CENTURY INFERRED FROM CENTRAL PACIFIC CORAL RECORDS Diane M. Thompson, Julia E. Cole, Alexander W. Tudhope

Abstract

Due to the far-reaching effects of tropical Pacific sea-surface temperatures (SSTs) on global climate, accurate regional climate projections require an understanding of the influence of anthropogenic climate change on the background condition and variability in the tropical Pacific. However, global climate models and instrumental climate data disagree regarding observed and future trends within the basin. Paired Sr/Ca and

records from corals may be used to reconstruct SST and (salinity), extending the limited instrumental dataset and offering an independent means to assess the thermal and hydrological response to climate change. However, a limited number of long paired

(Sr/Ca and coral records, particularly from the equatorial central Pacific, currently limits the ability to assess low-frequency variability or long-term trends in the

background state. Here, we present new replicated Sr/Ca-SST, , and records from the Republic of Kiribati (Onotoa and Maiana atolls) and central Line Islands

(Jarvis island) and use these records along with the published record from Palmyra atoll, northern Line Islands [Nurhati et al. 2011] to assess 20th-century variability and trends in

SST and salinity in the central tropical Pacific. Further, we use records from these sites to assess changes in the E-W SST gradient across the date line (between 173°E and

160°W) and the N-S SST gradient between the equator and 5°N. We find that Sr/Ca-SST variations are highly reproducible between two cores from Kiribati, while weaker correspondence between the records from Jarvis island may be attributed to age model 69

uncertainty or diagenetic alteration of the replicate core, which will be addressed in future work. We limited our analysis to records that correlate to local and regional SSTs. We

find that the Sr/Ca-SST, , and reconstructions from Onotoa, Maiana and

Jarvis display strong interannual and low-frequency variability that is strongly related to the El Niño-Southern Oscillation (ENSO) and Pacific decadal SST variations, particularly that associated with eastern Pacific-type ENSO events and the Pacific

Decadal Oscillation (PDO). The coral records suggest that the central tropical Pacific becomes warmer and fresher during positive phases of the PDO, resulting in a weakening of the E-W and N-S gradients. Superimposed on this interannual to decadal variability in these records is a general warming and freshening trend at these sites. Along with the record from Palmyra atoll (6ºN), these records suggest a slight strengthening of the E-W

SST gradient and little change in the N-S gradient over the 20th century. Finally, although

the coral reconstructions suggest that the tropical Pacific has generally freshened, the salinity response to climate change may be spatially heterogenous. Taken together, these results are consistent with a ramp up of the global hydrological cycle and a strengthening of the zonal SST gradient in response to anthropogenic warming.

1. Introduction

The state of the tropical Pacific Ocean plays a large role in global climate through its impact on atmospheric circulation [e.g., Rasmussen and Carpenter 1982; Trenberth

1998; Seager et al. 2005; Liu and Alexander 2007; Bulić et al. 2012; Graf and Zanchettin 70

2012]. Thus, accurate regional climate projections require an understanding of the influence of anthropogenic climate change on the background condition and variability in the tropical Pacific. Despite uniform warming to the basin, variability and/or the strength of the zonal and meridional gradients may change in response to changes in the strength of the Walker circulation [Vecchi et al. 2006; Bunge and Clarke 2009] and upwelling in the eastern Pacific [Clement et al. 1996] or along the equator [Liu et al. 2005; DiNezio et al. 2009; Karnauskas and Cohen 2012]. Changes in ENSO-related variance may in turn influence the background state within the basin through changes in the frequency and spatial pattern of ENSO events. For example, the recent prevalence of ENSO events with maximum anomalies in the central Pacific, so called El Niño Modoki or central Pacific

(CP) type ENSO, suggests that ENSO anomalies may become more strongly tied to the central Pacific [Yeh et al. 2009].

However, detection of change in the tropical Pacific over the past century has been limited by the sparse nature of historical observations and the presence of strong natural variability on interannual to decadal timescales. As a result, available SST data products disagree regarding the trend over the historical period [e.g., Vecchi et al. 2008,

Deser et al. 2010]. Coupled global climate models (CGCMs) similarly disagree regarding trends over the historical period [Thompson et al. 2011; Yeh et al. 2012] and under future anthropogenic warming, although the models generally suggest a weakened zonal SST gradient under continued anthropogenic forcing [Meehl et al. 2007; Guilyardi et al. 2012]. 71

By extending the limited historical observations, high-resolution coral paleoclimate records have greatly improved our understanding of interannual [Cole et al.

1993; Tudhope et al. 2001; Evans et al. 2002; Cobb et al. 2003] to decadal variability and trends [Urban et al. 2000; Cobb et al. 2001; Ault et al. 2009; Nurhati et al. 2009; 2011a] within the tropical Indo-Pacific. Numerous studies have demonstrated that the stable oxygen isotope (18O) composition of the coral skeleton is a robust proxy for variability within the basin as it is related to the temperature and isotopic composition of seawater

18  Osw (and thus salinity) at the time of growth [Fairbanks et al. 1997]. This feature of

18 coral  O records is advantageous for reconstructing past ENSO variability from corals

[e.g., Cole et al. 1993], as the atmosphere and ocean are tightly coupled on interannual

18 timescales [Bjerknes 1969]; however, changes in  Osw may confound the interpretation of low-frequency temperature variability and trends. In contrast, the ratio of Sr to Ca in the coral skeleton has been shown to be a more strict proxy for SST variability [Beck et al. 1992; Alibert and McCulloch 1997], although some work suggests that it may be particularly susceptible to growth-related effects [Grove et al. 2013] and diagenetic alterations [e.g., Sayani et al. 2011]. Nonetheless, capitalizing on the dual dependence

18 18 of  O on temperature and  Osw, a number of studies have demonstrated that robust

18 reconstructions of temperature and  Osw variability can be developed from paired Sr/Ca and 18O records [e.g., McCulloch et al. 1994; Gagan et al. 1998; Hendy et al. 2002;

Kilbourne et al. 2004; Linsley et al. 2006; Cahyarini et al. 2008]. Such Sr/Ca-SST and

18  Osw reconstructions from corals may be used to assess both the thermal and 72

hydrological response of the tropical Pacific to climate change [e.g., Nurhati et al. 2009;

2011a].

Despite the potential for paired Sr/Ca-18O records from corals to reconstruct SST and salinity variability within the basin, there are still a limited number of records long enough to assess low-frequency variability and trends within the basin. Further, long

18 Sr/Ca-SST and  Osw reconstructions have only been developed for one site within the central to eastern Pacific [Nurhati et al. 2011a], a critical region for assessing trends in zonal gradients and ENSO-related variability. Nurhati et al. [2011a] demonstrate that strong interannual variability associated with both classical eastern Pacific (EP) and CP

18 type ENSO events are recorded in SST and  Osw reconstructed from coral Sr/Ca and

18O from Palmyra atoll in the central tropical Pacific (6ºN, 162ºW). Consistent with previous studies [Urban et al. 2000; Cobb et al. 2001; Ault et al. 2009], they also find strong decadal variability in both records. They find that SSTs reconstructed from coral

Sr/Ca at this site is strongly tied to central Pacific variability associated with CP type

ENSO events and the North Pacific gyre oscillation (NPGO) [Di Lorenzo et al. 2008],

18 while reconstructed  Osw is more strongly tied to eastern Pacific variability associated with ENSO and the Pacific Decadal Oscillation [Mantua et al. 1997]. Superimposed on the strong variability observed at this site, they also find a warming and freshening trend that is consistent with a ramp-up of the global hydrological cycle [Held and Soden 2006] and a slowdown of the Walker circulation within the basin [e.g., Vecchi et al. 2006];

In this work we use replicated records from the Republic of Kiribati and the central Line Islands to assess 20th-century variability and trends in SST and salinity in the 73

central tropical Pacific. When considered together along with the previously published record from Palmyra Atoll (6ºN, 162ºW , [Nurhati et al. 2009; 2011a], these new records can be used to address the spatial fingerprint of warming within the basin in terms of changes in the E-W SST gradient across the dateline and the N-S equatorial SST gradient. The differential impact of eastern Pacific and central Pacific type ENSO events at these sites also allows us to address trends in these two types of ENSO events.

2. Study area

We present replicated coral 18O and Sr/Ca records from the Republic of Kiribati

(Onotoa and Maiana Atoll) and the central Line Islands (Jarvis Atoll) (Table B-1). These sites span ~25 degrees longitude across the equatorial Pacific, and therefore differ not only in their mean conditions, but also in their response to ENSO-related interannual variability (Figure B-1). The current E-W gradient across the dateline between Jarvis and the Republic of Kiribati (Onotoa and Maiana Atolls) is approximately 1.5-2ºC, while the

N-S gradient between Palmyra (at 5ºN) and Jarvis at the equator is ~1ºC. These gradients weaken during most El Nino events, as Jarvis warms more than Onotoa or Palmyra during a “canonical” EP El Nino event. Typical EP ENSO anomalies at Jarvis are around

1.8 ºC, compared to anomalies of ~0.4-0.5ºC at the other sites. In contrast, these sites warm a comparable amount during a CP ENSO event, with anomalies of around 0.9 to

1.4 ºC (Figure B-1).

3. Methods

3.1 Coral sampling

Cores of massive Porites spp. corals were collected from shallow reefs offshore of Maiana (1995), Onotoa (2003) and Jarvis (1999) atolls using an underwater hydraulic drill (Table B-1). The resulting cores were halved and slabbed for further analysis. X- ray images of the slabs were used to visualize the annual growth bands and to identify optimal sampling paths along the apex of corallite fans down the primary growth axis of the coral (Figures B-2-4). Further examination of the coral slabs under high magnification was used to assess centimeter-scale architectural structure, particularly the angle of the corallites with respect to the core. Samples were collected along transects running parallel to the growth direction of individual corallites, and areas of suboptimal sampling, where corallites grew at an angle relative to the coral slab [DeLong et al.,

2013], were noted. Coral powders were drilled continuously at 1 mm increments along 7 mm wide sampling paths down the Onotoa and Jarvis 1 coral slabs using an automated

Sherline CNC benchtop vertical milling system to obtain approximately monthly resolved records (given typical annual extension rates of 1-2 cm). Maiana coral slabs were drilled manually for 18O analysis [Urban et al. 2000]; in sections where resampling of these slabs was necessary, samples were drilled with the automated milling system parallel to the original sampling path of Urban et al. [2000] (Figure B-3).

Scanning electron microscope (SEM) images of samples from the top and bottom of Jarvis 2 were used to assess the presence of alteration or diagenesis of the coral material. SEM images of the bottom of the Jarvis 2 core were targeted on regions of 75

unusual geochemical behavior and reveal that in these regions, the surface of the coral aragonite has been altered (Figure B-5). The presence of large cold (higher Sr/Ca and

18O) anomalies in this section of core is consistent with the presence of secondary material [Allison 1996; Enmar et al. 2000; Allison et al. 2007; Sayani et al. 2011].

Extreme cold anomalies in the Jarvis 2 coral record around 1869-1872 and 1898 were excluded from all statistical analysis. Further, trend analyses presented here focus on the

20th-century portion of the coral record, which shows no sign of alteration (Figure B-5).

3.2 Geochemical analysis

3.2.1 Elemental analysis

Elemental analyses utilized a JY Optima 2C inductively coupled plasma atomic emission spectrometer (ICP-AES) at the University of Arizona. Splits (0.35-0.45 mg) of homogenized coral powder samples from each core were acidified in 2ml of 5% trace metal grade HNO3 to obtain a target dilution of the unknown sample solution of ~80 ppm calcium. We measured the ratio of strontium (407.77 nm) to calcium (393.37 nm) in these diluted coral samples on the ICP-AES. A reference solution was measured between every sample to remove low-frequency drift in the ratios observed throughout individual runs, following the method outlined by Schrag [1999]. Approximately every 60 samples throughout a run, we measured a dilution series of 5 gravimetrically certified high purity standards (HPS) to correct for variations in the Sr/Ca ratio with Ca concentration [Schrag

1999]. We analyzed replicates of an in-house coral standard solution along with these matrix standards to monitor long-term instrumental drift. The long-term analytical 76

precision (1σ) was 0.0305 mmol/mol for the high purity standards (0.33% RSD) and

0.044 mmol/mol for the in-house coral standard (0.49% RSD). The average method detection limit (MDL) of the ICP-AES was 0.029 mmol/mol. The MDL was calculated based on 7 replicate coral powder samples bracketed by blanks and the 5 HPS:

= where t is the Student’s t-value at the 99% confidence level for six degrees of freedom

(n-1, t=3.14) and s is the standard deviation of Sr/Ca ratios in the seven replicates [U.S.

Environmental Protection Agency, 1997].

3.2.2 Stable isotope analysis

Stable isotope analyses of samples from Jarvis were completed on a VG ISOCARB automated preparation system integrated with a VG ISOGAS PRISM II isotope ratio mass spectrometer at the Scottish Universities Environmental Research Centre (SUERC).

All other isotopic analyses were performed on a VG Optima Dual Inlet isotope ratio mass spectrometer with a common acid bath at the University of Arizona (Onotoa) or

University of Colorado (Maiana; [Urban et al. 2000]). Analytical precision of the 18O measurements reported here from all instruments is ± 0.08 ‰ (1σ).

3.3 Chronology development

The age models for these records were based on the identification of the seasonal cycle and major historical El Niño events in the 18O, 13C and Sr/Ca series. The chronology for the Jarvis 1 core was based solely on the Sr/Ca variations, as 18O and 77

13C measurements were only obtained at ~annual resolution for this core. Linear interpolation between the age model tie points was used to obtain monthly geochemical time series. Minor adjustments (≤ 1 year) to the original chronology were made based on comparison of the resulting geochemical series to HadISST1.1 SSTs (1871-1999).

3.4 Sr/Ca-SST calibration

SST observations from the coral sites are virtually absent for most of the 20th century

[Woodruff et al. 2005]. Due a lack of in situ temperature data from our coral sites and an overall limited set of historical temperature observations for the central tropical Pacific, we use gridded SST products to develop our Sr/Ca-SST calibration. We test the sensitivity of the Sr/Ca-SST calibration (Figures B-6-8) to the choice of SST products:

HadISST [Rayner et al., 2003], HadISST3 [Kennedy et al. 2011]), ERSSTv2 [Smith and

Reynolds, 2004], ERSSTv3b [Smith et al., 2008], NCEP reanalysis [Kalnay et al., 1996], and IGOSS [Reynolds et al. 2002]). As previously suggested [Corrège, 2006], we found that the choice of SST dataset had a large effect on the slope of the Sr/Ca-SST relationship (x axis, Figures B-6-8). For example at Onotoa, the slope ranged from -

0.0399 to -0.0683 mmmol/mol ºC-1 using OLS and -0.0696 to -0.0992 mmmol/mol ºC-1 using RMA over the 1950-1999 calibration interval. Similarly large ranges of slopes were observed at Maiana and Jarvis across SST products (Maiana: -0.0337 to -0.0648 mmmol/mol ºC-1 for OLS and -0.0763 to -0.114 mmmol/mol ºC-1 for RMA over the

1950-1993 calibration interval; Jarvis: -0.0434 to -0.0539 mmmol/mol ºC-1 for OLS and -

0.0739 to -0.0917 mmmol/mol ºC-1 for RMA over the 1950-1998 calibration interval). 78

We also explored the effect of different pre-processing strategies, including monthly raw, monthly anomalies and annual series and found similar results among resolutions and between raw and anomaly series. The only exception to this was that of annual series over the short calibration period beginning in 1982, as the sample size of the calibration set was very small. Finally, we utilized different calibration periods (beginning in 1982-,

1950-, and core start date; and end dates corresponding to either the common period ending in 1993 or the full period covered by each record). We settled on an interval that begins in 1950 when data coverage expands to obtain the largest potential calibration sample size; calibration results were not sensitive to the end date. Finally, we explored diverse methods (weighted least squares [WLS], ordinary least squares [OLS], and reduced major axis regression [RMA]).

The coral-based Sr/Ca-SST reconstructions presented here were generated by calibrating the Sr/Ca record from each site with SST from the nearest 2º x 2º gridbox of

ERSSTv3b dataset [Smith et al., 2008] using weighted least squares (WLS) linear regression [Thirumalai et al. 2011]. Following the maximum likelihood estimation

(MLE) method of York et al. [2004], this method accounts for the uncertainty in the regression estimate associated with the analytical precision of the Sr/Ca measurements

(1σ: 0.0305 mmol/mol) and the estimated error standard deviation of the SSTs. The

WLS method was chosen for calibration, as errors in Sr/Ca and SST may produce considerable bias in the regression estimate with OLS regression [Solow and Huppert

2004]. Although results from calibration with ERSSTv3b SSTs are presented here so that absolute SST values could be reconstructed, the variability and trends of reconstructed 79

Sr/Ca-SST were very similar when the HadiSSTv3 SST anomaly data [Kennedy et al.

2011] were used. The other data products do not include error estimates, which are required by the WLS regression method. Nonetheless, we also calculate the mean and standard deviation of the slope of the Sr/Ca-SST relationship obtained using RMS and

WLS with all SST products and intervals tested (N=14) to test the sensitivity of the results to the calibration. OLS regression results were excluded from this analysis, as they do not account for error in both datasets, which may produce significant slope biases

[Solow and Huppert 2004].

18 3.5  Osw reconstruction

18 18 We reconstruct  Osw from the paired Sr/Ca and  O records at each site by

18 removing the Sr/Ca-derived SST contribution to the  Ocoral values, following the centering method of Cahyarini et al. [2008]:

̅̅ ̅̅ ̅̅̅̅̅̅̅̅ 푙 ̅̅̅̅̅̅̅̅ = ( 푐표𝑟𝑎푙 푐표𝑟𝑎푙) (푆 푎 푆 푎) 푙

where is the anomaly calculated from the 푐표𝑟𝑎푙 and Sr/Ca

anomalies, 푙 is the regression slope of the 푐표𝑟𝑎푙-SST relationship, and 푙 is the regression slope of the Sr/Ca- SST relationship. This method differs from that of the

18 original methodology to reconstruct  Osw developed by McCulloch et al. [1994] and

Gagan et al. [1998] in that it removes the influence of the intercept of the 18O-SST and

Sr/Ca-SST relationships, which are susceptible to site-specific offsets. This method 80

produces similar results to that of Ren et al. [2002], which uses the first derivative of the

18 Sr/Ca and  O records to remove the intercepts. We define 푙 as -0.2 ± 0.03 ‰/ºC based on the experimental and theoretical relationship between and temperature observed in carbonates and particularly well-studied coral genera [e.g., Epstein et al., 1953; Evans et al., 2000; Juillet-Leclerc and Schmidt, 2001; Ren et al., 2002; Lough et al., 2004]. We define the slope of the Sr/Ca-SST relationship ( 푙 for each site based on the local calibration between Sr/Ca and ERSSTv3b SSTs as described above.

3.6 Uncertainty estimation

Uncertainty in the absolute values and trends reported here are calculated based on the contribution of analytical precision, error in the Sr/Ca-SST and -SST relationships and error in the trend regression estimate [after Nurhati et al. 2011b].

Uncertainty in the absolute values of 푐표𝑟𝑎푙 is based on analytical precision, and the

uncertainty in the 푐표𝑟𝑎푙 trends is calculated by quadratically combining the uncertainty associated with analytical precision and the trend estimate:

2 2 = √ +

where is the (1σ) analytical precision (0.08 ‰) and is the slope error of the coral

trend. Error in absolute Sr/Ca-SST values are calculated from the analytical precision of the Sr/Ca measurements ( 𝑟 𝑎 , error in the intercept ( 𝑎 and slope ) of the Sr-Ca-SST calibration, and the mean Sr/Ca (푆̅̅ ̅̅ ̅ ̅̅푎̅) as follows: 81

2 2 2 𝑟 𝑎 = √ 2 + ( 푆̅̅ ̅̅ ̅ ̅̅푎̅) [( ) + ( ) ] 𝑎 푆̅̅ ̅̅ ̅ ̅̅푎̅

As the intercept of the Sr/Ca-SST relationship does not affect the Sr/Ca-SST trend error, the Sr/Ca-SST trend error is calculated by quadratically combining the error from analytical precision ( 𝑟 𝑎 in units of ºC), the Sr/Ca-SST slope ( , and the trend estimate ( as follows:

2 2 2 𝑟 = √ + + 𝑟 𝑎 𝑎

where the uncertainty in the trend associated with the Sr/Ca-SST slope ( represents the difference between the SST trends calculated with the minimum and maximum

Sr/Ca-SST slopes. The error in the SST variability is even smaller, as reconstructed

Sr/Ca-SST variability is only affected by the error from analytical precision (with the error from the slope and intercept contributing equally to all values in the record).

Uncertainty in the trends in the Sr/Ca-SST gradient between two sites is calculated from their joint errors and the error in the trend estimate ( :

2 2 2 2 2 𝑟 = √ + + + + 𝑟 𝑎 𝑟 𝑎 𝑎

Finally, uncertainty in the values reconstructed from centered coral and

Sr/Ca-SST is calculated from the analytical precision of both the Sr/Ca and

measurements ( 𝑟 𝑎 and in units of ‰) and the slope of the -SST and

Sr/Ca-SST relationships 푙 푙) [Cahyarini et al. 2008]: 82

2 2 푙 2 = √ + ( ) 𝑟 𝑎 푙

The uncertainty of the trend is calculated by quadratically combining the errors associated the analytical precision ( 𝑟 𝑎 and in units of ‰), the Sr/Ca-SST slope

( , the -SST slope ( the and the trend estimate ( :

2 2 2 2 2 = √ + + + + 𝑟 𝑎

As above, the uncertainty in the trend associated with the Sr/Ca-SST slope ( and

-SST slope ( represents the difference between the trends calculated with the

minimum and maximum slopes. Finally, the error in variability is obtained from combining the error associated with analytical precision of the Sr/Ca and measurements.

To assess the effect of individual years on the trends reported here, we also perform a trend sensitivity analysis. One hundred trend analyses were performed on each record, with two years randomly removed for each iteration. The mean and standard devation of the trend magnitude in the 100 iterations is reported here for comparison.

The relatively large uncertainty values (3.5-5°C) associated with reconstructing absolute SSTs reported here are consistent with other studies [Nurhati et al. 2011] and reflect the cumulative error associated with analytical precision and the Sr/Ca-SST relationship (slope and intercept). In contrast, uncertainties in reconstructing relative 83

variability and trends are much smaller (~0.3-0.4°C) and support the use of these data to address 20th-century changes in trend and variability. All uncertainties are listed in Table

B-3.

4. Results

4.1 Replication of coral Sr/Ca and variations

We use replicate cores from the Republic of Kiribati (Onotoa and Maiana atolls) and Jarvis island to test the reproducibility of the observed Sr/Ca and variations.

Onotoa and Maiana atoll are separated by about 400 km, and the cores from Jarvis island were collected ~100 meters apart. All cores were collected from sites bathed by open ocean conditions to minimize the influence of reef-scale processes. We found strong agreement between the Sr/Ca and records from Onotoa and Maiana atolls (Figure

B-9), with significant positive correlations observed between all records from the two sites (Table B-2) and absolute mean values that agree within error (Table B-3). The

and records from Maiana displayed stronger negative anomalies than

Onotoa during many strong El Niño events, particularly those of the late 20th century

(e.g., 1982-83 and 1991-92), suggesting a stronger ENSO-related hydrological response at this site. The average absolute difference between reconstructed Sr/Ca-SSTs at Onotoa and Maiana is 0.58ºC over the 1912-1993 period of overlap, which is less than the uncertainty in the reconstructed Sr/Ca-SSTs at these two sites (Table B-3), and within 2σ of the analytical precision of the Sr/Ca measurements in terms of SST (0.77 ºC). SST 84

observations also suggest that some of this difference may be attributed to a true mean

SST difference between these sites (e.g., 0.2 ºC difference between Onotoa and Maiana in

IGOSS SSTs). The remainder of the observed Sr/Ca-SST difference between these sites

(0.378ºC) is within the 1σ error of the Sr/Ca measurements (0.386 ºC).

Despite the close proximity of the two Jarvis records, the agreement among these records is relatively weak. Nonetheless, the Jarvis records (Figure B-10) are significantly correlated to each other (Table B-2), with an average absolute difference in reconstructed

Sr/Ca-SSTs of 2.06 ºC over the 1867-1999 period of overlap (or 1.92ºC over the 1912-

1993 period). Although considerably larger than the difference between the Kiribati coral records, this difference is still within error of the absolute Sr/Ca-SST values (Table

B-3).

The Jarvis 1 record also displayed little correspondence with the Kiribati coral records and local and regional SST indices (Tables B-2, B-4 & B-5), suggesting that a climate signal may not be reliably reconstructed from the Jarvis 1 coral record. The low agreement between the Jarvis 1 record and other regional records may be attributed in part to large uncertainty in the age model for this record which stems from a lack of high- resolution and measurements from this core, which are still currently in development. Seasonal cycles in isotope records from these equatorial sites greatly aid age model development, as corals from these sites display only faint annual density banding and moderate Sr/Ca seasonality. The possibility of secondary alteration, even in such a young coral, also must be assessed using SEM images [e.g., Sayani et al. 2011].

All further analyses exclude the Jarvis 1 record until these issues can be resolved. 85

4.2 Coral Sr/Ca-SST reconstruction

Strong correlations between SST and Sr/Ca measured in the Onotoa, Maiana and

Jarvis 2 cores supports the interpretation of these Sr/Ca records in terms of local SST variability (Table B-4). Calibrating these Sr/Ca records with local ERSSTv3b SSTs over the period of best historical data coverage (1950-) using WLS regression [York et al.

2004, Thirumalai et al. 2011] to reconstruct absolute Sr/Ca-SST values, we obtain the following relationships between SST and coral Sr/Ca:

푆푆푇 𝑎𝑟2 = 푆 푎 𝑎𝑟2 (Jan 1950- Dec 1998, r= -0.56)

푆푆푇 𝑎 𝑎 𝑎 = 푆 푎 𝑎 𝑎𝑎 (Jan 1950-Dec 1993, r= -0.543)

푆푆푇 표 표𝑎 = 푆 푎 표 표𝑎 (Jan 1950-Dec 1999, r=-0.67)

The slope of the Sr/Ca-SST relationship at these sites falls in the upper end of the range of slopes that have been reported for Porites sp. corals [Corrège, 2006]. This may be attributed to the choice of WLS regression to account for error in both the Sr/Ca and SST measurements, as the slope of the Sr/Ca-SST relationship was greater at all sites when the

WLS method is used over the commonly applied OLS regression (Figures B-6-8).

The strength of the Sr/Ca-SST trend at these sites is sensitive to the slope of the

Sr/Ca-SST relationship, which differs depending on the SST product chosen (Figures B-

6-8). To test the influence of the Sr/Ca-SST relationship at each site on the observed variability and trends, we calculated the mean and standard deviation of the Sr/Ca-SST slope at each site. Using results from the RMS and WLS regression of all SST products and all time periods analyzed (N=14), we find a mean and standard deviation of the slope 86

of -0.089±0.0104 mmol/mol ºC-1 at Onotoa, -0.095±0.0157 mmol/mol ºC-1 at Maiana, and -0.0919±0.0179 mmol/mol ºC-1 at Jarvis.

4.3 Coral , Sr/Ca-SST and variability

The geochemical records from Jarvis and Kiribati display strong interannual to decadal variability tied to ENSO and low-frequency variations within the Pacific basin.

The , Sr/Ca-SST and records from Onotoa, Maiana and Jarvis 2 all display significant negative correlations with eastern Pacific (Niño 3.4) SSTs, consistent with warm, wet conditions at these sites during El Niño events (Table B-5). With the

exception of Onotoa , these records also are significantly negatively correlated with the El Niño Modoki index (EMI, [Ashok et al. 2007]. However, the correlation with the EMI was weaker than with Niño 3.4 SSTs in all cases, suggesting a stronger response to eastern Pacific ENSO events at these sites (or stronger variability associated with those

events). The Sr/Ca and records from Kiribati, and the and records from Jarvis, correlate positively with the NPGO index [Di Lorenzo et al. 2008], and all

records (with the exception of Onotoa ) correlate negatively with indices for the

PDO [Mantua et al. 1997; Kaplan et al. 2000]

4.4 20th-century trends

Over the 1912-1993 interval common to the records, the Sr/Ca-SST reconstructions from Onotoa, Maiana, and Jarvis islands indicate that significant warming occurred at these sites. Onotoa warmed by 1.35±0.385ºC, Maiana by 0.656±0.281ºC, and 87

Jarvis by 0.537±0.44ºC between 1912 and 1993. Using the mean Sr/Ca-SST slope, we find similar warming trends of 1.2±0.34ºC at Onotoa, 0.75±0.32ºC at Maiana, and

0.4±0.33ºC at Jarvis. All 3 coral records also display a significant trend towards more negative (depleted) values over this interval, consistent with warming and/or freshening (Table B-3). The negative trend observed in the Maiana and Jarvis 2 records is larger than can be explained by (Sr/Ca-derived) SST alone, suggesting that

freshening has contributed to the at trend at these sites. The series reconstructed from the paired Sr/Ca and records at Maiana and Jarvis thus display significant negative trends of -0.306 ± 0.098 ‰ and -0.228 ± 0.119 ‰, respectively.

These trends suggest a freshening of ~1.13 ± 0.098 PSU at Maiana and 0.84 ± 0.119 PSU

at Jarvis Island, based on the empirical relationship between and SSS for the tropical Pacific [Fairbanks et al. 1997; LeGrande and Schmidt 2006]. In contrast, no

significant trend in was observed at Onotoa Atoll.

The Sr/Ca-SST reconstructions from the Onotoa and Maiana suggest a warming of ~0.7-1.4ºC (±0.3-0.4ºC), which is greater than the warming of ~0.5 ºC (±0.4ºC) observed further east at Jarvis Island. As a result, the coral Sr/Ca-SST reconstructions suggest either no change or a slight strengthening of the E-W SST gradient across the dateline of ~0.2-0.8ºC (± 0.5-0.6 ºC) between 1912 and 1993 (Figure B-11). In contrast, the warming trend observed at Jarvis island (~0.5±0.4ºC) is comparable to that observed at Palmyra Atoll (~0.4 ±0.26ºC) [Nurhati et al. 2009; 2011a], suggesting no significant change in the N-S equatorial SST gradient over the 1912-1993 period (-0.12± 0.63ºC;

Figure B-12). Calculating the gradient trends and the error estimate of the trend from the 88

mean and error of potential calibration slopes at each site, we find that the gradient trends were not sensitive to the calibration slope uncertainty. The coral Sr/Ca-SST reconstructions still suggest no change or a slight strengthening of the E-W SST gradient of ~0.4-0.8ºC (± 0.5ºC) (Figure B-13) between 1912 and 1993, and no significant change in the N-S equatorial SST gradient (-0.014± 0.4ºC; Figure B-14) over this period. Note that in all cases, the sensitivity of the trend magnitude to the years included in the analysis was far less than the overall trend uncertainty (Table B-3), suggesting that these results are not sensitive to the time period covered analysis.

Capitalizing on the strong coherence between the Onotoa and Maiana Sr/Ca-SST records, we produce a composite record for the Republic of Kiribati covering the full

20th-century. This record suggests a warming of 1 ± 0.5ºC between 1900 and 1998, compared to a warming of 0.79± 0.4ºC at Jarvis. Consistent with the 1912-1993 trend, these records suggest a slight strengthening of the E-W SST gradient across the dateline, although the uncertainty does not preclude a stable gradient (-0.21± 0.65ºC) (Figure B-

15). The trends at Palmyra (0.23± 0.26ºC) and Jarvis (0.79± 0.4ºC) over the full 20th- century show a slight weakening of the N-S SST gradient of -0.55± 0.51ºC (Figure B-16).

In no case do we reconstruct a weakening E-W gradient across the dateline over either the shorter common interval or the longer (full-century) interval.

When combined with other published records, the reconstructions from this work suggest that the salinity trend is highly patchy in the tropical Pacifc (Figure B-

17). We find that the magnitude of the trend at these sites is sensitive to the

Sr/Ca-SST calibration method utilized, which varies greatly among published records 89

(Figures B-18-20). Here, we apply a common calibration approach to all published records (WLS regression against ERSSTv3b SSTs over the 1972-1992 interval common to all records), and find considerable spatial variability remains in the magnitude of the

trend over the 1972-1992 period (Figure B-17).

5. Discussion

This study presents three Sr/Ca and records from the central Line Islands

and the Republic of Kiribati, expanding the network of coral-based SST and reconstructions from the equatorial tropical Pacific. We use these replicated records to assess (5.1) the reproducibility of coral Sr/Ca measurements from these sites, (5.2) the relationship between Sr/Ca variations and local SST and the sensitivity of the SST reconstruction to calibration methodology, (5.3) the connection between variations in the coral records and regional climate, and (5.4) climatic implications of 20th-century trends

in SST, and equatorial Pacific gradients.

5.1 Reproducibility of coral Sr/Ca

Recent work has brought into question the reproducibility of coral Sr/Ca variations among closely spaced coral cores [e.g. Linsley et al., 2006; Calvo et al., 2007;

Pfeiffer et al. 2009; Grove et al. 2013]. The amplitude of variability [Pfeiffer et al.

2009], ENSO signal, and long-term trends [Grove et al. 2013] may differ between replicate Sr/Ca records. These discrepancies among replicate coral records may be 90

attributed to differences in the growth response of the individual corals to changes in SST

[Felis et al., 2003; 2004; Grove et al. 2013]. Nonetheless, application of a dendrochronology approach to replicate cores from the Great Barrier Reef [Hendy et al.

2002] and New Caledonia [DeLong et al. 2007; 2013] has demonstrated highly reproducibility among cross-dated cores, with differences in reconstructed Sr/Ca-SST across replicate New Caledonia cores that were less than the analytical uncertainty

[DeLong et al. 2007; 2013].

In this work, we further tested the reproducibility of coral Sr/Ca measurements from replicate cores from Jarvis and Kiribati (Onotoa and Maiana atolls). Despite being separated by a distance of ~400 km, we find strong agreement between the Maiana and

Onotoa Sr/Ca records, with absolute values that differ by an average of only 0.58ºC. This difference is much less than the uncertainty of the reconstructed absolute Sr/Ca-SST values themselves (Table B-3) and well within the range expected from analytical uncertainty and real SST differences. The replicate Sr/Ca records from Jarvis Atoll were also significantly, but weakly, correlated with one another, with an average absolute difference of ~2ºC in the reconstructed Sr/Ca-SST. The weak correspondence between replicate records from Jarvis Island may be attributed to chronological uncertainty in the

Jarvis 1 record; subannually resolved and records are currently being developed to improve the age model for this record.

5.2 Sr/Ca-SST relationship

The reported relationship between Sr/Ca and SST varies greatly among published records, with slopes ranging from -0.04 to -0.09 mmol/mol ºC-1 and a mean value around

-0.06 mmol/mol ºC-1 [Corrège 2006 and references within; Nurhati et al. 2009; DeLong et al. 2007]. Much of this discrepancy may be due to differences in the calibration approach employed in these studies, particularly with respect to the SST data used to calibrate the Sr/Ca measurements [Corrège 2006]. Consistent with this previous work, the Sr/Ca-SST relationship that we derive here is sensitive to the calibration approach.

The Sr/Ca-SST relationship is sensitive to the choice of SST product and to the choice of the regression method (Figures B-6-8). Independent of the SST product selected, we find that OLS regression consistently returns shallower slopes of the Sr/Ca-SST relationship than either RMA or WLS methods (both of which account for error in SST and Sr/Ca and produce robust estimates of the uncertainty of the slope). Although the application of

RMA regression is becoming more widespread [e.g., Shen and Dunbar 1995; Quinn et al.

1998; Quinn and Sampson, 2002; DeLong et al. 2007; Nurhati et al. 2009; 2011a], most published Sr/Ca-SST reconstructions have used OLS regression for the calibration of the

Sr/Ca thermometer. The slope of the Sr/Ca-SST relationships reported here for Jarvis,

Onotoa and Maiana coral records (-0.069 to -0.109 mmol/mol ºC-1), calculated from WLS regression against ERSSTv3b, therefore lie on the upper end of the reported calibration equations from the literature [see Corrège 2006], but are similar to that reported by

Nurhati et al. [2009] using RMA regression. 92

These results emphasize a need for a unified approach to calibrating Sr/Ca measurements within the coral paleoclimate community, and suggest that using the mean of the range of potential calibration slopes in the absence of local in situ SST data may improve the agreement among records. If a single method is chosen, the Sr/Ca-SST slopes for two different records may fall on opposite ends of the fairly wide range of potential values, leading to differences in the reconstructed records. We find that the difference in Sr/Ca-SST trend over the 1912-1993 period between Onotoa and Maiana is reduced using the mean calibration approach, such that the trends are within error of each other when the mean calibration slope is used.

5.3 Coral , Sr/Ca-SST and variability

The geochemical records from Jarvis and Kiribati display strong interannual to decadal variability tied to ENSO and low-frequency Pacific variations, as observed in the

Palmyra atoll records [Nurhati et al. 2011a]. However, in the Palmyra records, SST variability is tied to central Pacific variability, while salinity variability correlates with the eastern Pacific. Our new records correlate to variability in both the eastern and central

tropical Pacific. With the exception of the reconstruction from Onotoa Atoll, all of our records significantly correlate with Niño 3.4 SST, the ENSO Modoki index (EMI), the PDO and the NPGO (Table B-5). These records clearly track the full range of ENSO- related and low-frequency variability among the modes that influence these sites (Figure

B-1 & B-21, Table B-7). These reconstructions provide additional support for significant 93

decadal variability in the tropical Pacific, as observed in other [e.g. Ault et al. 2009] and Sr/Ca [e.g. DeLong et al. 2007; Nurhati et al. 2011a] records from the basin.

In contrast to Palmyra, where Sr/Ca-SST reconstructions correlate best with central Pacific variability, our geochemical records are more strongly related to eastern

Pacific variability, as indicated by the stronger correlations with Niño 3.4 SSTs and the

PDO. Our data suggest that the central tropical Pacific is warmer and fresher during positive phases of the PDO, as observed from 1925-1946 and 1977-1998 [Mantua et al.

1997], although the coral reconstructions suggest greater uncertainty in the decadal salinity response (Tables B-5 & B-7). The absence of a strong NPGO signature in our coral records may be attributed to the weak influence of this mode on SST variability at our sites; the PDO influence is stronger at our sites (Figure B-21). As a

result, the change in Sr/Ca, and associated with changes in the PDO are larger than observed with changes in the NPGO (Table B-7).

Overall variability is larger at Jarvis than at the Kiribati sites, because it is more sensitive to variability emanating from eastern Pacific upwelling. This larger variability is evident in the geochemical records, where regression slopes (the change associated with changes in the indices) tend to be steeper for Jarvis (Table B-7). The Jarvis SST reconstruction suggests that the EP ENSO events of 1983 and 1986 were particularly strong at this site, with anomalies of up to 4.2ºC. These anomalies are much larger than observed in the IGOSS SST dataset (~2.5ºC). Anomalies of similar magnitude (4.3 ºC) in the record and the absence of alteration in the top section of core, support the interpretation of these anomalies as an ENSO-related reduction in upwelling at this site. 94

Further, secondary aragonite typically found altered material results in anomalies in the opposite direction (resulting in a cold bias) [e.g., Sayani et al. 2011]. These years also displayed below average growth rate (10-14 mm/year vs. mean of 20 mm/year), as commonly observed during ENSO-related heat stress [see Lough and Cooper 2011 and reference therein]. Nonetheless, there is no evidence of nonlinearity in the Sr/Ca-SST relationship, suggesting that the growth response to these warm anomalies did not impact the Sr/Ca-SST calibration.

As a result of the strong fingerprint of the PDO at Jarvis, the E-W and N-S SST gradients also display strong low-frequency variability tied to the PDO. Both the E-W and N-S gradients are weaker during a positive phase of the PDO (1925-1946 and 1977-

1998) (Figures B-11-16, Table B-6), as Jarvis warms relative to Onotoa, Maiana and

Palmyra. An abrupt shift in the strength of the SST gradients was observed with the shift from positive to negative PDO in 1946. The E-W and N-S SST gradients also weaken during a positive phase of the NPGO (e.g., 1975-1979, 1987-1990) (Figures B-11-16,

Table B-6), but to a much lesser degree.

5.4 Coral-derived trends in tropical Pacific SST, and gradients

5.4.1 Coral-derived SST trends

Recent climate modeling work suggests that anthropogenic warming may be enhanced in the central Pacific [Xie et al. 2010] and along the equator [Liu et al., 2005;

DiNezio et al. 2009; Gastineau and Soden, 2009; Widlansky et al., 2013]. However, the available instrumental data products disagree with respect to the magnitude and even sign 95

of the temperature trends throughout the basin, particularly in the eastern e quatorial

Pacific. While enhanced warming in the central and eastern Pacific is observed in the

ERSSTv3b dataset, the HadISST v1 dataset displays cooling in the central and eastern

Pacific. As a result, the ERSSTv3b dataset suggests a weakening of the equatorial E-W

SST gradient across the dateline (between 173°E and 160°W), while the HadISST v1 dataset suggests a strengthening of the gradient driven by cooling in the east. Further, in contrast to a weakening of the N-S SST gradient expected as a result of enhanced equatorial warming, both datasets display a strengthening of the N-S SST gradient between the equator and 6°N.

To address these uncertainties in the response of the E-W and N-S SST gradients to warming, we use the Sr/Ca-SST reconstructions from Jarvis Atoll (160ºW) along with those from the Republic of Kiribati (173-175ºE) and Palmyra Atoll (6ºN) to reconstruct changes in the E-W and N-S SST gradients, respectively. These coral records suggest a strengthening of the E-W temperature gradient of ~0.2-0.8 (±0.52-0.58) ºC and no change in the N-S equatorial temperature gradient (-0.012 ± 0.63 ºC) over the 1912-1993 period common to all of the coral records. Although the strength of the warming trend at individual sites was sensitive to the slope of the Sr/Ca-SST relationship (Figures B-6-8), we find similar results in the E-W and N-S gradients over the 1912-1993 period when the mean Sr/Ca-SST relationship was used at each site. Nonetheless, using a stacked record from the Kiribati to assess trends over the full 20th-century (1900-1998), we find that a signature of enhanced equatorial warming emerges from the noise of the reconstructed

Sr/Ca-SST trends, while the observed strengthening of the E-W SST gradient is not 96

significant over the 20th-century. Additional records from the central tropical Pacific are needed to confirm the significance of these trends. Moreover, our records end in the

1990s, and assessing the full extent of Pacific trends in a warming world requires that we bring these records up to present with new cores.

Nonetheless, in no case was a weakening of the E-W equatorial SST gradient implicated by the coral Sr/Ca-SST reconstructions from these sites. Thus, although all sites displayed significant warming trends, the pattern of warming was most similar to that observed in the HadISST dataset, with warming enhanced in the west relative to the east. This apparent strengthening of the E-W temperature gradient across the dateline between Jarvis Island and the Republic of Kiribati if significant may be driven by a change in the background state in the basin, either due to a westward migration of the western Pacific warm pool (WPWP) [Delcroix and Picaut 1998; Liu and Huang, 2000;

Huang and Liu, 2001; Cravatte et al., 2009] associated with a weakening of the equatorial trade winds [e.g., Tokinaga et al. 2012] or through the ocean thermostat mechanism [Clement et al. 1996], by which upwelling counters the warming in the eastern portion of the basin.

On the other hand, the change in the E-W gradient may be associated with a change in ENSO itself, specifically an increase in CP ENSO events [Yeh et al. 2009], whose anomalies are centered near Onotoa and Maiana. To address whether increased

CP ENSO events could be causing the increasing gradient, we analyzed the variance of the high-pass filtered records (using an 11-year normalized hamming window) in 20-year moving windows (Figure B-22). As Jarvis experiences large temperature anomalies 97

during both EP and CP ENSO events, its reconstructed ST variance is higher than that at

Kiribati. The Jarvis record displays lower variance during the 1920-60s, a period when

ENSO was known to be relatively weak from other ENSO-sensitive records. After the

1960s, the variance of the Jarvis record increases dramatically in response to the large

ENSO events observed during the late 20th century. In contrast, Onotoa and Maiana display very little change in interannual variability over this period. Therefore, the E-W gradient trend cannot be attributed to an increase in CP ENSO events, as there is no evidence for an increase in CP variance relative to the east in the coral Sr/Ca-SST reconstructions.

5.4.2 Coral-derived trends

Available salinity products disagree regarding the magnitude and even direction of the trend over the instrumental period (Figure B-17). As with SST, these discrepancies may be attributed to the paucity of historical climate data from the

equatorial Pacific. Here, we use reconstructed from paired coral Sr/Ca and to assess salinity trends within the basin. We find that although the coral reconstructions generally suggest that the basin is freshening in response to anthropogenic warming, the

magnitude of the response is highly spatially heterogeneous within the tropical

Pacific. Coral reconstructions from Maiana and Jarvis display significant trends towards lighter values, suggesting a freshening trend between 1912 and 1993, whereas no significant trend is observed at Onotoa Atoll. The freshening trends observed at Maiana and Jarvis are consistent with the wet-get-wetter pattern associated with the strengthening 98

of the global hydrological cycle [Held and Soden 2006] and the shift in precipitation associated with a weakened Walker circulation [Vecchi et al. 2006; Zhang and Song

2006; Bunge and Clarke 2009]. Increased advection of low salinity waters from the

WPWP [e.g. Delcroix and Picaut 1998; Liu and Huang, 2000; Huang and Liu, 2001;

Cravatte et al., 2009] may also contribute to the freshening trend at these sites.

We find that the magnitude of the trend at these sites is sensitive to the slope of the Sr/Ca-SST relationship (Figures B-18-20), which depends on the SST product and calibration methodology utilized. These results provide further emphasis for the need for a unified approach to Sr/Ca reconstruction within the coral paleoclimate community. Applying a unified approach to calibrating the available monthly Sr/Ca-SST records from the tropical Pacific (WLS regression of Sr/Ca against ERSSTv3 SSTs over

1972-1992 period common to all records), we find that although the reconstructions generally suggest freshening across the tropical Pacific consistent with the increase in the global hydrological cycle [Held and Soden 2006], the salinity trends are still spatially heterogeneous (Thompson et al. 2011; Figure B-17). Differences in analytical methods and the absence of a common standard among laboratories for the

measurement of Sr/Ca may contribute to the discrepancies among the resulting reconstructions [Corrège 2006; Hathorne et al. 2013]. On the other hand, given the heterogeneity of the salinity trend over the historical record (Figure B-17) and the extremely limited data coverage for long-term variability [Delcroix et al. 2011], the potential for a spatially heterogeneous salinity response to climate change cannot be dismissed and needs to be further investigated. The paucity of salinity observations that 99

contribute to these historical datasets adds considerable uncertainty to the assessment of historical salinity trends, emphasizing the need for additional paired coral Sr/Ca and

records from the tropical Pacific to address salinity trends. Additional coral

reconstructions may help constrain the anthropogenic fingerprint of salinity change within the basin.

6. Conclusions

We use replicated records from the Republic of Kiribati and the central Line

Islands to assess the reproducibility of Sr/Ca records from these sites and their relationship with SST. Paired with records from the same cores, we then

th reconstruct SST and and use these new records to assess 20 century variability and trends in SST and SSS in the central tropical Pacific. We find that Sr/Ca-SST variations are highly reproducible between two cores from Kiribati, despite being separated by over 400 km. After accounting for the observed difference in SST between these two sites (~0.2 ºC), the difference in reconstructed Sr/Ca-SSTs (0.58 ºC) is within the precision of the Sr/Ca measurements. Although we also find agreement between replicate records from Jarvis island, the agreement is much weaker than observed between the Kiribati records despite being separated by only ~100 meters, with a mean difference of ~2ºC in reconstructed Sr/Ca-SSTs. This difference may be attributed to uncertainties in the Jarvis 1 record, which also displays weak correspondence with local and regional SSTs. SEM imaging of the skeleton and high-resolution isotope records 100

from this core will be developed in future work to improve the age model for this record and screen this record for diagenesis.

The Sr/Ca-SST, , and records from Onotoa, Maiana and Jarvis 2 cores display strong interannual and low-frequency variability that is strongly related to

ENSO and Pacific decadal SST variations. Although the records display significant correlations to both central and eastern Pacific variability interannual and decadal timescales, the records are more strongly tied to eastern Pacific variability (EP ENSO and

PDO, respectively). The coral records suggest that the central tropical Pacific becomes warmer and fresher during positive phases of the PDO, resulting in a weakening of the E-

W and N-S gradients between these sites. The shift in the PDO from a positive to negative phase in 1946 is strongly recorded in the SST gradient between these sites. The

SST gradients also weaken significantly during a positive phase of the NPGO, but to a much lesser degree than associated with shifts in the PDO.

Superimposed on the variability in these records was a warming and freshening

trend, except at Onotoa where there was no significant trend. The Sr/Ca-SST reconstructions suggest that the warming at Kiribati may have been greater than further east at Jarvis, resulting in an apparent strengthening of the SST gradient. In contrast, the warming trend is similar in magnitude between Jarvis near the equator, and Palmyra at

6ºN, suggesting that there has not been a strong enhancement of warming along the equator and change in the N-S SST gradient. We find no clear trend in interannual variability within the central Pacific, suggesting that there has been no detectable increase

th in CP type ENSO events over the 20 century. Coral reconstructions from 101

Maiana and Jarvis suggest a freshening trend at these sites, while no significant

trend was observed at Onotoa. Along with other reconstructions from the tropical Pacific, these records suggest that the salinity response to climate change may be very patchy within the basin.

Our results are limited in that they end during the 1990’s, and some of the largest responses to a warming world may arguably be seen since that time. Moreover, our analysis of the gradient only considers a limited zonal extent across the dateline (172E –

160W). Additional records from equatorial sites in the easternmost Pacific would add greatly to this analysis, as would cores that extend to present from all sites. Additional paired Sr/Ca and records from corals would improve our understanding of the thermal and hydrological fingerprint of warming in the tropical Pacific.

Acknowledgements

This work was funded by NOAA CCDD program (grant # NA08OAR4310682 funded to

J.E. Cole), the University of Arizona Department of Geosciences and Institute of the

Environment, and the Philanthropic Educational Organization.

102

References Alibert, C., and M.T. McCulloch (1997), Strontium/calcium ratios in modern Porites

corals from the Great Barrier Reef as a proxy for sea surface temperature: Calibration

of the thermometer and monitoring of ENSO, Paleoceanography, 12(3), 345-363.

Allison, N. (1996), Geochemical anomalies in coral skeletons and their possible

implications for palaeoenvironmental analyses, Marine Chemistry, 55(3), 367-379.

Allison, N., A.A. Finch, J.M. Webster, and D.A. Clague (2007), Palaeoenvironmental

records from fossil corals: the effects of submarine diagenesis on temperature and

climate estimates. Geochimica et Cosmochimica Acta, 71(19), 4693-4703.

Ashok, K., S.K. Behera, S.A. Rao, H. Weng, and T. Yamagata (2007), El Niño Modoki

and its possible teleconnection, Journal of Geophysical Research, 112(C11).

Ault, T. R., J.E., Cole, M.N. Evans, H. Barnett, N.J. Abram, A.W. Tudhope, and B.K.

Linsley (2009), Intensified decadal variability in tropical climate during the late 19th

century, Geophysical Research Letters, 36(8), L08602.

Beck, J.W., R.L. Edwards, E. Ito, F.W. Taylor, J. Recy, F. Rougerie, P. Joannot, and C.

Henin (1992), Sea-surface temperature from coral skeletal strontium/calcium

ratios, Science, 257(5070), 644-647.

Bjerknes, J. (1969), Atmospheric teleconnections from the equatorial pacific 1, Monthly

Weather Review, 97(3), 163-172. 103

Bulić, I. H., Č. Branković, and F. Kucharski (2012), Winter ENSO teleconnections in a

warmer climate, Climate dynamics, 38(7-8), 1593-1613.

Bunge, L., and A.J. Clarke, (2009), A Verified Estimation of the El Niño Index Niño-3.4

since 1877, J. Clim., 22:3979-3992.

Cahyarini, S. Y., M. Pfeiffer, O. Timm, W.C. Dullo, and D.G. Schönberg (2008),

Reconstructing seawater δ18O from paired coral δ18O and Sr/Ca ratios: Methods, error

analysis and problems, with examples from Tahiti (French Polynesia) and Timor

(Indonesia), Geochimica et Cosmochimica Acta, 72(12), 2841-2853.

Calvo, E., J.F. Marshall, C. Pelejero, M.T. McCulloch, M.K. Gagan, and J.M. Lough

(2007), Interdecadal climate variability in the Coral Sea since 1708 A.D.,

Palaeogeography, Palaeoclimatology, Palaeoecology, 248, 190–201

http://dx.doi.org/10.1016/ j.palaeo.2006.12.003.

Carton, J.A., and B.S. Giese (2008), A reanalysis of ocean climate using Simple Ocean

Data Assimilation (SODA), Mon. Weather Rev., 136, 2999–3017,

doi:10.1175/2007MWR1978.1.

Carton, J.A., et al. (2000), A Simple Ocean Data Assimilation analysis of the global

upper ocean 1950-1995, Part 1: methodology, J. Phys. Oceanogr., 30, 294-309.

Clement AC, Seager R, Cane MA, and Zebiak SE, 1996. An Ocean Dynamical

Thermostat. Journal of Climate 9:2190-2196 104

Cobb, K. M., C.D. Charles, and D.E. Hunter (2001), A central tropical Pacific coral

demonstrates Pacific, Indian, and Atlantic decadal climate connections, Geophysical

Research Letters, 28(11), 2209-2212.

Cobb, K.M., C.D. Charles, H. Cheng, and R.L. Edwards (2003), El Niño-Southern

Oscillation and tropical Pacific climate during the last millennium, Nature,

424(6946), 271 – 276

Cole, J. E., R.G. Fairbanks, and G.T. Shen (1993). Recent variability in the Southern

Oscillation: Isotopic results from a Tarawa Atoll coral, Science, 260(5115), 1790-

1793.

Corrège, T. (2006), Sea surface temperature and salinity reconstruction from coral

geochemical tracers, Palaeogeography, Palaeoclimatology, Palaeoecology, 232(2),

408-428.

Cravatte, S., T. Delcroix, D. Zhang, M. McPhaden, and J. Leloup (2009), Observed

freshening and warming of the western Pacific Warm Pool, Climate Dynamics, 33,

565-589.

Delcroix, T., G. Alory, S. Cravatte, T. Corrège, and M.J. McPhaden (2011), A gridded

sea surface salinity data set for the tropical Pacific with sample applications (1950–

2008), Deep Sea Research Part I: Oceanographic Research Papers, 58(1), 38-48. 105

Delcroix, T., and J. Picaut (1998), Zonal displacement of the western equatorial Pacific

“fresh pool”, Journal of Geophysical Research: Oceans (1978–2012), 103(C1), 1087-

1098.

DeLong, K. L., T.M. Quinn, and F.W. Taylor (2007), Reconstructing twentieth‐century

sea surface temperature variability in the southwest Pacific: A replication study using

multiple coral Sr/Ca records from New Caledonia, Paleoceanography, 22(4).

DeLong, K.L., T.M. Quinn, F.W. Taylor, K. Lin, and C.-C. Shen (2012), Sea surface

temperature variability in the southwest tropical Pacific since AD 1649, Nature

Climate Change, 2(11), 799-804, doi:10.1038/nclimate1583.

DeLong, K.L., T.M. Quinn, F.W. Taylor, C.-C. Shen, K. Lin (2013), Improving coral-

base paleoclimate reconstructions by replicating 350 years of coral Sr/Ca variations,

Palaeogeography, Palaeoclimatology, Palaeoecology, 373, 6-24, ISSN 0031-0182.

Deser, C., A.S. Phillips, and M.A. Alexander (2010), Twentieth century tropical sea

surface temperature trends revisited, Geophysical Research Letters, 37(10).

Di Lorenzo, E., N. Schneider, K.M. Cobb, P.J.S. Franks, K. Chhak, A.J. Miller, J.C.

McWilliams, S.J. Bograd, H. Arango, E. Curchitser, T.M. Powell, and P. Rivière

(2008), North Pacific Gyre Oscillation links ocean climate and ecosystem

change. Geophysical Research Letters, 35(8). 106

DiNezio, P.N., A.C. Clement, G.A. Vecchi, B.J. Soden, B.P. Kirtman, and S.-K. Lee,

2009. Climate Response of the Equatorial Pacific to Global Warming, J. Clim.,22,

4873-4892.

Enmar, R., M. Stein, M. Bar-Matthews, E. Sass, A. Katz, and B. Lazar (2000).

Diagenesis in live corals from the Gulf of Aqaba. I. The effect on paleo-

oceanography tracers, Geochimica et Cosmochimica Acta, 64(18), 3123-3132.

Environmental Protection Agency (EPA). 40 CFR Part 136, APPENDEX B Revision

1.11

Evans, M. N., A. Kaplan, and M.A. Cane (2002), Pacific sea surface temperature field

reconstruction from coral δ18O data using reduced space objective analysis,

Paleoceanography, 17(1), 7-1.

Fairbanks, R. G., M.N. Evans, J.L. Rubenstone, R.A. Mortlock, K. Broad, M.D. Moore,

and C.D. Charles, (1997), Evaluating climate indices and their geochemical proxies

measured in corals, Coral Reefs, 16(1), S93-S100.

Felis, T., J. Pätzold, and Y. Loya (2003), Mean oxygen-isotope signatures in Porites spp.

corals: Inter-colony variability and correction for extension-rate effects, Coral Reefs,

22(4), 328– 336, doi:10.1007/s00338-003-0324-3.

Felis, T., G. Lohmann, H. Kuhnert, S. J. Lorenz, D. Scholz, J. Pätzold, S. A. Al-Rousan,

and S. M. Al-Moghrabi (2004), Increased seasonality in Middle East temperatures

during the last interglacial period, Nature, 429(6988), 164– 168,

doi:10.1038/nature02546 107

Gagan, M. K., L.K. Ayliffe, D. Hopley, J.A. Cali, G.E. Mortimer, J. Chappell, M.T.

McCulloch, and M.J. Head (1998). Temperature and surface-ocean water balance of

the mid-Holocene tropical western Pacific. Science, 279(5353), 1014-1018.

Gastineau, G., and B. J. Soden (2009), Model projected changes of extreme wind events

in response to global warming, Geophysical Research Letters, 36

Graf, H. F., and D. Zanchettin (2012), Central Pacific El Niño, the “subtropical bridge,”

and Eurasian climate, Journal of Geophysical Research: Atmospheres (1984–

2012), 117(D1).

Grove, C. A., S. Kasper, J. Zinke, M. Pfeiffer, D. Garbe‐Schönberg and G.J.A. Brummer

(2013), Confounding effects of coral growth and high SST variability on skeletal

Sr/Ca: implications for coral paleothermometry, Geochemistry, Geophysics,

Geosystems, 14, 4, 1277-1293.

Guilyardi, E., H. Bellenger, M. Collins, S. Ferrett, W. Cai, and A. Wittenberg (2012), A

first look at ENSO in CMIP5, Clivar Exchanges, 17(1), 29-32.

Hathorne, E. C., et al. (2013), Interlaboratory study for coral Sr/Ca and other element/Ca

ratio measurements. Geochemistry, Geophysics, Geosystems.

Held, I.M., and B.J. Soden (2006), Robust Responses of the Hydrological Cycle to

Global Warming. J. Clim., 19:5686-5699. 108

Hendy, E. J., M.K. Gagan, C.A. Alibert, M.T. McCulloch, J.M. Lough, and P.J. Isdale

(2002), Abrupt decrease in tropical Pacific sea surface salinity at end of Little Ice

Age, Science, 295(5559), 1511-1514.

Huang, B., and Z. Liu (2001), Temperature trend of the last 40 yr in the upper Pacific

Ocean. Journal of Climate, 14, 3738-3750.

Kalnay, E., et al. (1996), The NCEP/NCAR 40-Year Reanalysis Project. Bulletin of the

American Meteorological Society, 77(3), 437-471.

Kaplan, A., Y. Kushnir, and M.A. Cane (2000), Reduced Space Optimal Interpolation of

Historical Marine Sea Level Pressure: 1854-1992*, Journal of Climate, 13(16), 2987-

3002.

Karnauskas, K. B., and A.L. Cohen (2012), Equatorial refuge amid tropical

warming, Nature Climate Change, 2(7), 530-534.

Kennedy, J. J., N.A. Rayner, R.O. Smith, D.E. Parker, and M. Saunby (2011),

Reassessing biases and other uncertainties in sea surface temperature observations

measured in situ since 1850: 2. Biases and homogenization, Journal of Geophysical

Research: Atmospheres (1984–2012), 116(D14).

Kilbourne, K. H., T.M. Quinn, F.W. Taylor, T. Delcroix, and Y. Gouriou (2004), El

Nino–Southern Oscillation–related salinity variations recorded in the skeletal

geochemistry of a Porites coral from Espiritu Santo, Vanuatu,

Paleoceanography, 19(4), PA4002. 109

Linsley, B.K., A. Kaplan, Y. Gouriou, J. Salinger, P.B. deMenocal, G.M. Wellington, and

S.S. Howe (2006), Tracking the extent of the South Pacific convergence zone since

the early 1600s, Geochemistry, Geophysics, Geosystems, 7, Q05003

http://dx.doi.org/10.1029/2005GC001115.

Liu, Z., S. Vavrus, F. He, N. Wen, and Y. Zhong (2005), Rethinking Tropical Ocean

Response to Global Warming: The Enhanced Equatorial Warming*, Journal of

Climate, 18(22), 4684-4700.

Liu, Z., and M. Alexander (2007), Atmospheric bridge, oceanic tunnel, and global

climatic teleconnections, Reviews of Geophysics, 45(2).

Liu, Z., and B. Huang, (2000), Cause of tropical Pacific warming trend. Geophysical

Research Letters, 27, 1935-1938.

Lough, J. M., and T.F. Cooper (2011), New insights from coral growth band studies in an

era of rapid environmental change, Earth-Science Reviews, 108(3), 170-184.

Mantua, N.J., S.R. Hare, Y. Zhang, J.M. Wallace, and R.C. Francis (1997), A Pacific

interdecadal climate oscillation with impacts on salmon production, Bulletin of the

American Meteorological Society, 78, pp. 1069-1079.

McCulloch, M. T., M.K. Gagan, G.E. Mortimer, A.R. Chivas, and P.J. Isdale (1994), A

high-resolution Sr/Ca and δ18O coral record from the Great Barrier Reef, Australia,

and the 1982–1983 El Niño, Geochimica et Cosmochimica Acta, 58(12), 2747-2754.

Meehl, G.A., et al. (2007) Global Climate Projections. In: Climate Change 2007: The

Physical Science Basis. Contribution of Working Group I to the Fourth Assessment 110

Report of the Intergovernmental Panel on Climate Change. Cambridge University

Press, pp. 747-845. ISBN 9780521880091.

Nurhati, I.S., K.M. Cobb, C.D. Charles, and R.B. Dunbar (2009), Late 20th century

warming and freshening in the central tropical Pacific, Geophys. Res. Lett., 36,

L21606

Nurhati, I.S., K.M. Cobb, and E. Di Lorenzo (2011a), Decadal-scale SST and Salinity

Variations in the Central Tropical Pacific: Signatures of Natural and Anthropogenic

Climate Change. Journal of Climate, 24(13), 3294-3308. doi:

10.1175/2011JCLI3852.1.

Nurhati, I. S., K.M. Cobb, C.D. Charles, and R.B. Dunbar (2011b), Correction to “Late

20th century warming and freshening in the central tropical Pacific”, Geophysical

Research Letters, 38(24).

Pfeiffer, M., W.-C. Dullo, J. Zinke, and D. Garbe-Schönberg (2009), Three monthly coral

Sr/Ca records from the Chagos Archipelago covering the period of 1950–1995 A.D.:

Reproducibility and implications for quantitative reconstructions of sea surface

temperature variations, Int. J. Earth Sci., 98, doi:10.007/s00531-008-0326-z.

Quinn, T. M., T.J. Crowley, F.W. Taylor, C. Henin, P. Joannot, and Y. Join (1998), A

multicentury stable isotope record from a New Caledonia coral: Interannual and

decadal sea surface temperature variability in the southwest Pacific since 1657

AD, Paleoceanography, 13(4), 412-426. 111

Quinn, T. M., and D.E. Sampson (2002), A multiproxy approach to reconstructing sea

surface conditions using coral skeleton geochemistry, Paleoceanography, 17(4),

1062.

Rasmussen, E. M., and T.H. Carpenter (1982), Variations in tropical sea surface

temperature and surface wind ﬁelds associated with the southern oscillation/El

Niño, Mon. Weath. Rev, 110, 354-84.

Rayner, N. A., et al. (2003), Global analyses of sea surface temperature, sea ice, and

night marine air temperature since the late nineteenth century, J. Geophys. Res.,

108(D14), 4407, doi:10.1029/2002JD002670.

Ren, L., B.K. Linsley, G.M. Wellington, D.P. Schrag, and O. Hoegh-Guldberg (2002),

Deconvolving the δ18O seawater component from subseasonal coral δ18O and Sr/Ca at

Rarotonga in the southwestern subtropical Pacific for the period 1726 to

1997, Geochimica et cosmochimica acta, 67(9), 1609-1621.

Reynolds, R.W., N.A. Rayner, T.M. Smith, D.C. Stokes, and W. Wang (2002), An

Improved In Situ and Satellite SST Analysis for Climate, J. Climate, 15, 1609-1625.

Sayani, H. R., K.M. Cobb, A.L. Cohen, W.C. Elliott, I.S. Nurhati, R.B. Dunbar, K.A.

Rose, and L.K. Zaunbrecher (2011), Effects of diagenesis on paleoclimate

reconstructions from modern and young fossil corals, Geochimica et Cosmochimica

Acta, 75(21), 6361-6373. 112

Seager, R., N. Harnik, W.A. Robinson, Y. Kushnir, M. Ting, H.P. Huang, and J. Velez

(2005), Mechanisms of ENSO‐forcing of hemispherically symmetric precipitation

variability, Quarterly Journal of the Royal Meteorological Society, 131(608), 1501-

1527.

Schrag, D. P. (1999), Rapid analysis of high‐precision Sr/Ca ratios in corals and other

marine carbonates, Paleoceanography, 14(2), 97-102.

Shen, G. T., and R.B. Dunbar (1995), Environmental controls on uranium in reef

corals, Geochimica et Cosmochimica Acta, 59(10), 2009-2024.

Smith, T.M. and R.W. Reynolds (2004), Improved extended reconstruction of SST

(1854–1997), J. Climate, 17, 2466–2477.

Smith, T.M., et al. (2008), Improvements to NOAA's historical merged land-ocean

surface temperature analysis (1880-2006), J. Climate, 21, 2283-2296.

Solow, A. R., & Huppert, A. (2004). A potential bias in coral reconstruction of sea

surface temperature. Geophysical research letters, 31(6).

Stephans, C. L., T.M. Quinn, F.W. Taylor., and T. Corrège (2004), Assessing the

reproducibility of coral‐based climate records, Geophysical research letters, 31(18).

Thirumalai, K., A. Singh, and R. Ramesh (2011), A MATLAB™ Code to Perform

Weighted Linear Regression with (correlated or uncorrelated) Errors in Bivariate

Data, Journal Of The Geological Society of India, 77(April), 377-380. 113

Thompson, D. M., T.R. Ault, M.N. Evans, J.E. Cole, and J. Emile-Geay (2011),

Comparison of observed and simulated tropical climate trends using a forward model

of coral δ18O. Geophysical Research Letters, 38(14).

Tokinaga, H., S.P. Xie, A. Timmermann, S. McGregor, T. Ogata, H. Kubota, and Y.M.

Okumura (2012), Regional Patterns of Tropical Indo-Pacific Climate Change:

Evidence of the Walker Circulation Weakening*, Journal of Climate, 25(5), 1689-

1710.

Trenberth, K. E., G.W. Branstator, D. Karoly, A. Kumar, N.C. Lau, and C. Ropelewski

(1998), Progress during TOGA in understanding and modeling global teleconnections

associated with tropical sea surface temperatures, Journal of Geophysical Research:

Oceans (1978–2012), 103(C7), 14291-14324.

Tudhope, A. W., C.P. Chilcott, M.T. McCulloch, E.R. Cook, J. Chappell, R.M. Ellam,

D.W. Lea, J.M. Lough, and G.B. Shimmield (2001), Variability in the El Niño-

Southern Oscillation through a glacial-interglacial cycle, Science, 291(5508), 1511-

1517.

Urban, F. E., J.E. Cole, and J.T. Overpeck (2000), Influence of mean climate change on

climate variability from a 155-year tropical Pacific coral record, Nature, 407(6807),

989-993.

Vecchi, G.A., B.J. Soden, A.T. Wittenberg, I.M. Held, A. Leetmaa, and H.J. Harrison

(2006), Weakening of the tropical Pacific atmospheric circulation due to

anthropogenic forcing, Nature, 441, 73-76. 114

Vecchi, G.A. and B.J. Soden (2007), Global Warming and the Weakening of the

Tropical Circulation, J. Clim., 20, 4316-4340.

Vecchi, G.A., A. Clement, and B.J. Soden (2008), Examining the Tropical Pacific’s

Response to Global Warming, EOS, 89, 9, 81-83.

Widlansky, M. J., et al., 2013: Changes in South Pacific rainfall bands in a warming

climate. Nature Climate Change, 3, 417–423.

Woodruff, S. D., H.F. Diaz, S.J. Worley, R.W. Reynolds, and S.J. Lubker (2005), Early

ship observational data and ICOADS, Climatic Change, 73(1-2), 169-194.

Yeh, S.-W., J.-S. Kug, B. Dewitte, M.-H. Kwon, B.P. Kirtman, and F.-F. Jin (2009), El

Niño in a changing climate, Nature, 461, 511-514.

Yeh, S. W., Y.G. Ham, and J.Y. Lee (2012), Changes in the Tropical Pacific SST Trend

from CMIP3 to CMIP5 and Its Implication of ENSO*, Journal of Climate, 25(21),

7764-7771.

York, D., N.M. Evensen, M.L. Martínez., and J.D.B. Delgado (2004), Unified equations

for the slope, intercept, and standard errors of the best straight line, American Journal

of Physics, 72, 367.

Zhang, M., and H. Song (2006), Evidence of deceleration of atmospheric vertical

overturning circulation over the tropical Pacific, Geophys. Res. Lett., 33, L12701.

115

Figure B-1: (a) Average SSTs over the 1900-2011 period from the ERSST v3b dataset, (b) Composite eastern Pacific El Niño (EP) pattern, calculated as the average of the DJF SST anomaly from 5 EP events (1965–66, 1972–73, 1976–77, 1982–83, 1997–98), (c) same as in (b) for central Pacific El Niño (CP) pattern, calculated from 6 CP events (1968–69, 1990–91, 1994–95, 2002–03, 2004–05, 2006–07 [Ashok et al. 2007). The ‘x’s’ mark the locations of our coral samples from (west to east): Maiana Atoll, Onotoa Atoll, and Jarvis Island, and the circle denotes the location of the Palmyra Atoll coral record [Cobb et al., 2003; Nurhati et al., 2009; Nurhati et al. 2011a]. The mean SST at the coral sites was: 27.24ºC at Jarvis; 28.66ºC at Maiana; 28.82ºC at Onotoa; and 28.03ºC at Palmyra. EP ENSO events were associated with anomalies of: 1.79ºC at Jarvis; 0.42ºC at Maiana; 0.41ºC at Onotoa; and 0.54ºC at Palmyra. In contrast, CP ENSO events were associated with anomalies of: 1.38 ºC at Jarvis; 1.23ºC at Maiana; 1.14ºC at Onotoa; and 0.91ºC at Palmyra.

116

Figure B-2: Positive X-ray images of Porites sp. coral slabs from two cores collected offshore of Onotoa Atoll, Republic of Kiribati in 2003. Geochemical analyses were performed on samples from core 6 milled at 1 mm increments along transects following optimal growth features (indicated in yellow). Analyses of samples from core 5 (right) were used to patch the mill break in core 6 that occurred during collection. 117

Figure B-3: Positive X-ray images of Porites sp. coral slabs from core 2-3 collected at Maiana Atoll, Republic of Kiribati in January, 1995. Geochemical analyses were performed on samples previously milled at 1 mm increments along the paths indicated in yellow (Urban et al. 2000). Where necessary, additional powders were drilled parallel to the original sampling path. 118

Figure B-4: Positive X-ray images of Porites sp. coral slabs from core 99-1-2 collected offshore of Jarvis Atoll in 1999. Geochemical analyses were performed on samples from milled at 1 mm increments along transects following optimal growth features (indicated in yellow). Sampling ended where the coral’s growth axis changed such that no clear banding pattern was observed (with corallites oriented perpendicular to the slab). 119

Figure B-5: Comparison of Jarvis 2 surfaces with and without alteration. (a-d) Scanning Electron Microscope (SEM) images of unaltered surfaces near the top of the core and (e- h) SEM images of altered surfaces near the bottom of the core 120

Figure B-6: Sensitivity of the 20th century Onotoa Sr/Ca-SST trend (1900-1998) to the Sr/Ca-SST calibration using the available SST products (symbols) and Ordinary Least Squares (OLS), Reduced Major Axis (RMA), or Weighted Linear Regression (WLS) regression. For each product, the regression was performed with monthly anomalies over the 1982-1999 and 1950-1999 periods (NCEP & IGOSS: 1982-1999 only). Similar results were also obtained with raw and annual series and over the full period of overlap.

121

Figure B-7: Sensitivity of the 20th century Maiana Sr/Ca-SST trend (1912-1994) to the Sr/Ca-SST calibration using the available SST products (symbols) and Ordinary Least Squares (OLS), Reduced Major Axis (RMA), or Weighted Linear Regression (WLS) regression. For each product, the regression was performed with monthly anomalies over the 1982-1993 and 1950-1993 periods (NCEP & IGOSS: 1982-1993 only). Similar results were also obtained with raw and annual series and over the full period of overlap.

122

Figure B-8: Sensitivity of the 20th century Jarvis 2 Sr/Ca-SST trend (1900-1998) to the Sr/Ca-SST calibration using the available SST products (symbols) and Ordinary Least Squares (OLS), Reduced Major Axis (RMA), or Weighted Linear Regression (WLS) regression. For each product, the regression was performed with monthly anomalies over the 1982-1998 and 1950-1998 periods (NCEP & IGOSS: 1982-1998 only). Similar results were also obtained with raw and annual series and over the full period of overlap.

123

Figure B-9: Onotoa (solid) and Maiana (dotted) 18O, Sr/Ca-SST, and 18Osw. Trends over 1900-2000 and the 1912-1994 period of overlap are shown in black. Sr/Ca-SST was reconstructed from Sr/Ca and the local Sr/Ca-ERSSTv3b WLS calibration.

Figure B-10: Jarvis 2 (solid) and Jarvis 1 (dotted) 18O, Sr/Ca-SST, and 18Osw. Trends over the full record (1850-1999 for Jarvis 2 and 1867-1999 for Jarvis 1) and the 1912- 1994 period of overlap among coral records are shown in black. Sr/Ca-SST was reconstructed from Sr/Ca and the local Sr/Ca-ERSSTv3b WLS calibration. Altered areas of the Jarvis 2 record (highlighted in gray) were excluded from all analyses presented here. 124

Figure B-11: (top) Sr/Ca-SST reconstructed from Onotoa (red), Maiana (red dashed) and Jarvis 2 coral Sr/Ca (using the ERSSTv3b WLS calibration). (bottom) SST gradient across the date line calculated as the difference between reconstructed SST from Jarvis 2 and Onotoa and Maiana Atolls, Republic of Kiribati. The trend lines suggest an increasing E-W gradient.

125

Figure B-12: (top) Sr/Ca-SST reconstructed from Jarvis 2 (blue) and Palmyra (red) using the local Sr/Ca-ERSSTv3b WLS calibration. (bottom) N-S SST gradient calculated as the difference between Palmyra and Jarvis 2 Sr/Ca-SSTs, with the trend line indicating an insignificant decrease in the N-S gradient.

126

Figure B-13: (top) Sr/Ca-SST anomalies reconstructed from Onotoa (red), Maiana (red dashed) and Jarvis 2 coral Sr/Ca using the mean slope of the Sr/Ca-SST relationship from sensitivity analysis. (bottom) SST gradient across the date line calculated as the difference between reconstructed SST from Jarvis 2 and Onotoa and Maiana Atolls, Republic of Kiribati. The trend lines suggest an increasing E-W gradient.

127

Figure B-14: (top) Sr/Ca-SST anomalies reconstructed from Jarvis 2 (blue) and Palmyra (red) using the mean slope of the Sr/Ca-SST relationship from sensitivity analysis. (bottom) N-S SST gradient calculated as the difference between Palmyra and Jarvis 2 Sr/Ca-SSTs, with the trend line indicating no significant change in the N-S gradient.

128

Figure B-15: (top) Sr/Ca-SST reconstructed from the stacked Republic of Kiribati records (red), and Jarvis 2 coral Sr/Ca over the 1900-1998 period (using the ERSSTv3b WLS calibration). (bottom) SST gradient across the date line calculated as the difference between reconstructed SST from Jarvis 2 and the Kiribati stack. The trend line suggests an insignificant increase in the E-W gradient.

129

Figure B-16: (top) Sr/Ca-SST anomalies reconstructed from Jarvis 2 (blue) and Palmyra (red) over the 1900-1998 period (using the ERSSTv3b WLS calibration). (bottom) N-S SST gradient calculated as the difference between Palmyra and Jarvis 2 Sr/Ca-SSTs, with the trend line indicating a slight weakening of the N-S gradient. 130

18 Figure B-17: Trends over the 1972-1992 period in  Osw reconstructed from paired coral Sr/Ca and 18O (symbols) and historical SSS from (a) Simple Ocean Data Assimilation (SODA) version 2.1.6 [Carton and Giese, 2008], (b) Carton GOA Beta 7 18 [Carton et al., 2000], and (c) Delcroix [Delcroix et al. 2011].  Osw was reconstructed from corals at (from east to west): Christmas [Nurhati et al. 2009], Fanning [Nurhati et al. 2009], Jarvis island (this work), Palmyra [Nurhati et al. 2009; 2011a], Onotoa atoll (this work), Maiana atoll (this work & Urban et al. 2000], Vanuatu [Kilbourne et al. 2004], and New Caledonia [DeLong et al. 2012]. 131

th 18 Figure B-18: Sensitivity of the 20 century Onotoa  Osw trend (1900-2000) to the Sr/Ca-SST calibration using the available SST products (symbols) and Ordinary Least Squares (OLS), Reduced Major Axis (RMA), or Weighted Linear Regression (WLS) regression. For each product, the regression was performed with monthly anomalies over the 1982-1999 and 1950-1999 periods (NCEP 1982-1999 only). Similar results were also obtained with raw and annual series and over the full period of overlap (not shown). 132

th 18 Figure B-19: Sensitivity of the 20 century Maiana  Osw trend (1912-1994) to the Sr/Ca-SST calibration using the available SST products (symbols) and Ordinary Least Squares (OLS), Reduced Major Axis (RMA), or Weighted Linear Regression (WLS) regression. For each product, the regression was performed with monthly anomalies over the 1982-1993 and 1950-1993 periods (NCEP 1982-1993 only). Similar results were also obtained with raw and annual series and over the full period of overlap (not shown). 133

th 18 Figure B-20: Sensitivity of the 20 century Jarvis 2  Osw trend (1900-1998) to the Sr/Ca-SST calibration using the available SST products (symbols) and Ordinary Least Squares (OLS), Reduced Major Axis (RMA), or Weighted Linear Regression (WLS) regression. For each product, the regression was performed with monthly anomalies over the 1982-1998 and 1950-1998 periods (NCEP 1982-1998 only). Similar results were also obtained with raw and annual series and over the full period of overlap (not shown).

134

Figure B-21: Regression of SST (ERSSTv3b) [Smith et al., 2008] with indices of Pacific decadal variability: (a) Pacific Decadal Oscillation (PDO) index from 1900 to 2012 [Mantua et al. 1997] and (b) Negative NPGO index from 1950 to 2012 [Di Lorenzo et al. 2008]. The ‘x’s’ mark coral locations (west to east): Maiana Atoll, Onotoa Atoll, and Jarvis Island, and the circle denotes Palmyra Atoll [Cobb et al., 2003; Nurhati et al., 2009; Nurhati et al. 2011a]. The slope of the regression (change in temperature per unit change in the index) at the coral sites is 0.326 ºC for the PDO and 0.22 ºC for the - NPGO at Jarvis; 0.168 ºC for the PDO and 0.156 ºC for the -NPGO at Maiana; 0.178 ºC for the PDO and 0.134 ºC for the -NPGO at Onotoa; and 0.139 ºC for the PDO and 0.177 ºC for the -NPGO at Palmyra

135

Figure B-22: (top) Variance of the high pass filtered Sr/Ca-SST records from Onotoa (red), Maiana (red dashed) and Jarvis 2 (blue) in 20 year moving windows. (bottom) E-W difference in high-frequency variance.

136

Table B-1: Coral sites from which Sr/Ca records were developed in this study. The length and resolution of 18O records previously developed from these same coral cores are also noted along with the original reference (where applicable).

Site Latitude Longitude Coral 18O Reference Kiribati/Gilberts Maiana Atoll 0°55' N 173°02'E 1840-1995; 6/yr Urban et al. core 95-2-3 2000 Onotoa Atoll 1°52'S 175°34'E 1870-2002; 12/yr cores 03-5 & 03-6 Jarvis Island Core 99-1-2 0°022'S 160°02'W ~2 yr Core 99-2-4 0°022'S 159°59'W 1850-1999; 10/yr

137

Table B-2: Correlation between coral geochemical anomaly records. Anomalies were calculated for each series over the full record. Italcs, bold and bold italics denotes correlation coefficients (r) that were significant at the 95%, 99% and 99.9% confidence level, respectively.

Coral Onotoa Onotoa Maiana Maiana Jarvis 1 Jarvis 1 Jarvis 2 Jarvis 2 Palmyra Palmyra record Sr/Ca 18O Sr/Ca 18O Sr/Ca 18O* Sr/Ca 18O Sr/Ca 18O [Nurhati [Nurhati et al. et al. 2011] 2011] Onotoa 1 0.661 0.505 0.651 0.264 0.124 0.365 0.554 0.361 0.617 Sr/Ca Onotoa 18O 1 0.344 0.475 0.211 0.173 0.316 0.487 0.331 0.518 Maiana 1 0.691 0.209 0.262 0.328 0.528 0.335 0.493 Sr/Ca Maiana 18O 1 0.381 0.335 0.410 0.648 0.307 0.550 Jarvis 1 1 0.874 0.350 0.446 0.057 0.312 Sr/Ca Jarvis 1 1 0.315 0.542 0.056 0.465 18O* Jarvis 2 1 0.738 0.333 0.428 Sr/Ca Jarvis 2 18O 1 0.449 0.649 Palmyra 1 0.607 Sr/Ca [Nurhati et al. 2011] Palmyra 1 18O [Nurhati et al. 2011] *Annual series 138

Table B-3: Mean absolute values and the magnitude of the 20th century trend (1912- 18 18 1993) of coral Sr/Ca-derived SST,  O and (centered)  Osw using the ERSSTv3b WLS calibration and their associated uncertainty (1σ) calculated after Nurhati et al. [2011b]. Sr/Ca-SST trends calculated using the mean Sr/Ca-SST slope are also shown for comparison in parentheses. Mean and standard deviation of 100 iterations of the trend analysis, each with 2 years randomly removed from the record, are also shown to demonstrate the sensitivity of the trend analysis to the years included in the analysis.

18 18 Sr/Ca-SST (ºC)  O (‰)  Osw (‰) Mean absolute values Onotoa (1900-2002) 28.57 ± 4.63 -4.74 ± 0.08 6.8E-4± 0.08 Maiana (1912-1994) 28.57 ± 3.58 -4.57± 0.08 0.0027 ± 0.08 Jarvis 1 (1867.25-1999.25) 25.97 ± 12.07 -4.56 ± 0.08 0.0072± 0.08 Jarvis 2 (1850-1999.67) 27.07 ± 3.74 -4.63 ±0.08 -0.0342 ± 0.08

20th-century trends 1.35 ± 0.385 0.0844 ± 0.111 Onotoa -0.186 ± 0.08 (1.2 ± 0.34)

0.656 ± 0.280 Maiana -0.442 ± 0.08 -0.306 ± 0.098 (0.75 ± 0.32)

0.537 ± 0.442 Jarvis 2 -0.334 ± 0.08 -0.228 ± 0.119 (0.4 ± 0.33) -0.82 ± 0.59 Jarvis 2- Onotoa E-W gradient (-0.81 ± 0.48) -0.15 ± 0.52 Jarvis 2- Maiana E-W gradient (-0.37 ± 0.46) -0.12 ± 0.51 Palmyra-Jarvis 2 N-S gradient (0.014 ± 0.42) 20th-century trend sensitivity analysis Onotoa 1.354 ± 0.04 -0.185 ± 0.0096 0.0845 ± 0.006 Maiana 0.651 ± 0.026 -0.439 ± 0.014 -0.307 ± 0.009 Jarvis 2 0.541 ± 0.07 -0.336 ± 0.009 -0.229 ± 0.01 Jarvis 2- Onotoa E-W gradient -0.822 ± 0.068 Jarvis 2- Maiana E-W gradient -0.140 ± 0.026 Palmyra-Jarvis 2 N-S gradient -0.121 ± 0.07

139

Table B-4: Correlation between coral geochemical and local SST anomaly records. Anomalies were calculated for each series over the full record, and correlated with SST anomalies from the nearest grid box of HadISST (1871-1999.92; [Rayner et al., 2003]), HadISST3 (1854-2013.5, [Kennedy et al. 2011]), ERSSTv3b (1854-2008.92; [Smith et al., 2008]), and NCEP reanalysis (1981.83-2012; [Kalnay et al., 1996]) over the period of overlap. Italics, bold and bold italics denote correlation coefficients (r) that were significant at the 95%, 99% and 99.9% confidence level, respectively.

HadISST1 HadISST3 ERSSTv3b NCEP SSTa Record SSTa SSTa SSTa (1981.83-2012) (1871-1999.92) (1854-2013.5) (1854-2008.92) Onotoa (1900-2002) Sr/Ca -0.592 -0.512 -0.596 -0.740 18O -0.454 -0.397 -0.494 -0.450 Maiana (1912-1994) Sr/Ca -0.547 -0.386 -0.556 -0.648 18O -0.479 -0.392 -0.552 -0.672 Jarvis 1 (1867.25-199.25) Sr/Ca -0.324 -0.250 -0.332 -0.223 18O* -0.226 -0.228 -0.282 0.128 Jarvis 2 (1850-1999.67) Sr/Ca -0.486 -0.418 -0.488 -0.670 18O -0.494 -0.595 -0.507 -0.855 *Annual series

140

Table B-5: Correlation between coral geochemical and regional SST anomaly indices: Niño 3.4 SST anomalies, ENSO Modoki index [Ashok et al. 2007], North Pacific gyre oscillation (NPGO) index [Di Lorenzo et al. 2008], and the Pacific Decadal Oscillation (PDO) index from Mantua et al. [1997] and Kaplan et al. [2000]. Italics, bold and bold italics denote correlation coefficients (r) that were significant at the 95%, 99% and 99.9% confidence level, respectively.

Niño 3.4 SSTa Modoki Index NPGO index PDO index Kaplan PDO* Record (1871.04-1997.46) (1870-2012.42) (1950-2012.75) (2000-2012.75) (1857-1991)

Onotoa (1900-2002) Sr/Ca -0.504 -0.411 0.226 -0.315 -0.485 18O -0.530 -0.298 0.155 -0.273 -0.332 18  Osw -0.179 0.036 -0.0322 -0.033 0.129 Maiana (1912-1994)

Sr/Ca -0.471 -0.443 0.275 -0.241 -0.324 18O -0.526 -0.296 0.205 -0.279 -0.391 18  Osw -0.423 -0.141 0.101 -0.224 -0.325 Jarvis 2 (1850-1999.67) Sr/Ca -0.525 -0.369 0.0078 -0.379 -0.399 18O -0.550 -0.436 0.262 -0.275 -0.457 18  Osw -0.106 -0.150 0.329 0.102 -0.214

*Annual series

141

Table B-6: Correlation between SST gradient indices calculated from coral geochemical records and regional SST anomaly indices: Niño 3.4 SST anomalies, ENSO Modoki index [Ashok et al. 2007], North Pacific gyre oscillation (NPGO) index [Di Lorenzo et al. 2008], and the Pacific Decadal Oscillation (PDO) index from Mantua et al. [1997] and Kaplan et al. [2000]. Italics, bold and bold italics denote correlation coefficients (r) that were significant at the 95%, 99% and 99.9% confidence level, respectively.

Niño 3.4 SSTa Modoki Index NPGO index PDO index Kaplan PDO* Record (1871.04-1997.46) (1870-2012.42) (1950-2012.75) (2000-2012.75) (1857-1991)

E-W gradient Jarvis 2- Onotoa 0.310 0.058 0.108 0.206 0.225

Jarvis 2- Maiana 0.397 0.092 0.128 0.303 0.340 N-S gradient

Palmyra- Jarvis 2 -0.470 -0.112 -0.123 -0.405 -0.481 E-W variability gradient Jarvis 2- Onotoa 0.134 0.008 0.042 0.234 0.540

Jarvis 2- Maiana 0.144 0.111 -0.030 0.408 0.611

*Annual series

142

Table B-7: Regression between coral geochemical and regional SST anomaly indices: Niño 3.4 SST anomalies, ENSO Modoki index [Ashok et al. 2007], North Pacific gyre oscillation (NPGO) index [Di Lorenzo et al. 2008], and the Pacific Decadal Oscillation (PDO) index from Mantua et al. [1997] and Kaplan et al. [2000].

Niño 3.4 SSTa Modoki Index NPGO index PDO index Kaplan PDO* Record (1871.04-1997.46) (1870-2012.42) (1950-2012.75) (2000-2012.75) (1857-1991)

Onotoa (1900-2002) Sr/Ca (mmol/mol) -0.0443 -0. 0663 0.0139 -0.0204 -0.350 18O (‰) -0.133 -0.138 0.0272 -0.0508 -0.623 18  Osw (‰) -0.034 0.0126 -0.0041 -0.0046 0.170 Maiana (1912-1994)

Sr/Ca (mmol/mol) -0.0396 -0.0736 0.0220 -0.0148 -0.199 18O (‰) -0.175 -0.193 0.0633 -0.0666 -0.983 18  Osw (‰) -0.108 -0.0703 0.0242 -0.0411 -0.646 Jarvis 2 (1850-1999.67) Sr/Ca (mmol/mol) -0.0678 -0.0935 0.0008 -0.0349 -0.369 18O (‰) -0.195 -0.308 0.0633 -0.0664 -1.26 18  Osw (‰) -0.0276 -0.0758 0.0614 0.0195 -0.402

*Annual series

143

APPENDIX C

20th-CENTURY VARIABILITY IN TRADE WIND REVERSALS INFERRED FROM Mn/Ca IN A WESTERN PACIFIC CORAL

To be submitted to the professional journal: Geophysical Research Letters

144

20th-CENTURY VARIABILITY IN TRADE WIND REVERSALS INFERRED FROM Mn/Ca IN A WESTERN PACIFIC CORAL

Diane Thompson1, Julia Cole1,2, and Glen Shen3

1Department of Geosciences, University of Arizona

2Department of Atmospheric Sciences, University of Arizona

3Proposal Exponent, Seattle, WA

Abstract

Zonal wind strength and direction are fundamental components of the El Niño-Southern

Oscillation (ENSO). Historical observations of tropical Pacific winds are limited, and existing datasets disagree on long-term trends, emphasizing the need for independent data to assess the response of zonal winds to climate change. Earlier work using a 17-year dataset from Tarawa Atoll suggests that Mn/Ca in corals near west-facing lagoons record westerly winds associated with El Niño events. We compare a ~90 year Mn/Ca record from Tarawa with 20th-century reanalysis zonal wind and demonstrate a strong association between westerly winds and Tarawa coral Mn/Ca. The Mn/Ca record also captures several strong historical events that are underestimated or absent in the wind data. A reduction in Mn/Ca variability within this 90-year record suggests a decrease in westerly winds in the western equatorial Pacific. This reduction in Mn/Ca variability is consistent with a reduction in the frequency of westerly winds within the 20th century reanalysis, and suggests either a strengthening of the trade winds or a change in the frequency or location of ENSO-related westerly wind anomalies. Additional Mn/Ca

145 records extending to present are needed to replicate this record and quantify long-term trends.

1. Introduction

Global climate models predict slowdown of the Walker circulation with greenhouse gas forcing as a result of increased atmospheric stability and a reduced zonal

SST gradient [Vecchi and Soden 2007; Meehl et al. 2007]. However, detection of this expected weakening of the Walker circulation over the past century is limited by the short, imperfect observational record. Trends in reanalysis data sets [Feng et al. 2011; Li and Ren 2012], radiation fluxes [Chen et al. 2002], and sea level [Merrifield 2011;

Merrifield and Maltrud 2011] suggest a strengthening of the zonal winds and Walker circulation since the early 1990s. In contrast, historical sea level pressure, cloudiness, precipitation, mixed layer temperature, and thermocline tilt data suggest a reduction in zonal winds associated with a weakening of the Walker circulation [Wu and Xie 2003;

Vecchi et al. 2006; Power and Smith, 2007; Deser et al. 2010; Power and Kociuba 2011;

Tokinaga et al. 2012]. Some studies suggest that discrepancies among datasets may be attributed to biases in the reanalysis data caused by changes in the type, frequency and spatial resolution of the observations [e.g., Wu and Xie 2003; Tokinaga et al. 2012].

Discrepancies among historical wind data sets highlight the need for an independent method to assess zonal wind trends in the equatorial Pacific. The Mn/Ca ratio in corals from islands with west-facing lagoons may offer a proxy for westerly wind anomalies within the tropical Pacific [Shen et al., 1992] and thus provide clues regarding

146 changes in the zonal wind strength. Westerly winds preferentially occur prior to and during El Niño events, when the sea level pressure gradient is reduced, zonal winds are anomalously westerly and the tropical Pacific warm pool is extended eastward [e.g., Yu et al., 2003; Vecchi and Harrison, 2000; Eisenman et al., 2005; Seiki and Takayabu, 2007].

These wind bursts initiate downwelling Kelvin waves that propagate eastward and play an important role in the onset and maintenance of El Niño conditions [e.g., Vecchi and

Harrison 2000; Lengaigne et al. 2002; Federov 2002; Lengaigne et al. 2003; Eisenman et al. 2005; Seiki and Takayabu 2007]. Thus, westerly wind events are strongly tied to both the background state of the tropical Pacific (including zonal wind strength) and ENSO- related variability within the basin [e.g., Fedorov 2002; Seiki and Takayabu 2007].

Trends in the occurrence and strength of westerly winds may therefore provide insight into changes in both ENSO variability and mean state.

In a short coral core from Tarawa, Shen et al. [1992] found that Mn/Ca was high during El Niño events, when zonal winds within the tropical Pacific were weak, and conditions around Tarawa were warm and wet. They propose that westerly winds trigger strong physical mixing within the westward facing lagoon, releasing Mn from lagoonal sediments enriched in Mn relative to the overlying water column. This remobilized Mn, dissolved in seawater, is then incorporated into coral skeletons on the nearby fringing reef

(Figure C-1). The resolution of the coral records and mixing time of lagoonal waters onto the reef precludes the ability to reconstruct individual wind burst events that occur over a few days. Nonetheless the elevated concentration of Mn in seawater resulting from frequent westerly wind anomalies and bursts prior to and during El Niño events are captured in the Tarawa coral Mn/Ca record. Mn/Ca in corals from islands with west-

147 facing lagoons, such as Tarawa Atoll, may therefore provide a record of westerly wind anomalies associated with El Niño events [Shen et al., 1992] and improve our understanding of interannual variability and trends in tropical Pacific zonal winds.

Despite the promise of this method, the association between wind anomalies and coral Mn/Ca has not been applied beyond the 17-year coral record from Tarawa Atoll

[Shen et al., 1992]. The paucity of wind observations and the minute concentrations of

Mn in corals at sites remote from continental or volcanic influences (e.g. 18-47nmol/mol;

Shen et al., [1992]) has limited robust calibration and widespread application of this proxy. The recent development of the 20th-century reanalysis (20CR) data set [Compo et al., 2011] allows us to further test the relationship between coral Mn/Ca and westerly wind anomalies. In this work we use a long (~90 year) Mn/Ca record from Tarawa and

20CR zonal wind data to demonstrate a robust relationship between zonal wind anomalies and coral Mn/Ca. Comparison with coral δ18O from the same core and with

ENSO SST indices adds weight to the connection of Mn/Ca and El Niño conditions. We discuss implications for trends in zonal winds within the basin.

2. Tarawa coral record

Cores from a massive Porites sp. colony were collected in 1989-90 on the southwest fringe of Tarawa Atoll (Figure C-1, 1°N, 172°E). Cole et al. [1993] presented a monthly δ18O record from this material that correlates strongly with rainfall, SST, and large-scale ENSO indices. The current study uses the same core material and adopts the

δ13C-based age model from the earlier work. Here, we present a record of Mn/Ca

148 produced on a graphite furnace atomic absorption spectrophotometer (GFAAS) following the method of Shen and Boyle [1988] and Shen et al. [1992]. Due to the large sample sizes required to detect trace concentrations of Mn in the coral skeleton, a quarterly resolved record of Mn/Ca variations was developed from this core and compared to the monthly δ18O record. X-ray images of the annual bands were used to sample the core at approximately quarterly resolution using a miniature band saw. Although this sampling approach was ideal for capturing the Mn/Ca signal, it is important to note that the accuracy of this sampling method may result in some variability in the temporal coverage of the Mn/Ca data between years. Each sample was washed and acid-leached before being crushed, sieved and cleaned with ultrasonic agitation in peroxide, reducing media and acid solutions. 74-mg sample aliquots were then dissolved in HNO3, and cobalt-

APDC coprecipitation in 1.5 ml polyethelyne microcentrifuge vials was used to isolate the transition metals (see Shen and Boyle [1988] and Shen et al. [1991] for further details on the methodology). A Hitachi model Z-9000 GFAAS was then used to measure Mn/Ca in the dissolved sample aliquots.

3. Zonal wind

We use the suite of available data sets to assess zonal wind variability and trends around Tarawa. Consistent with previous work [e.g., Wu and Xie, 2003], we found considerable disagreement among the various wind products over the instrumental period in the 10° x 10° region centered on Tarawa atoll (4°S-6°N, 166°-176°E, Figure C-S1a).

Much of this uncertainty stems from the limited number of zonal wind observations surrounding Tarawa, particularly before 1950 (Figure C-2b). Although these products

149 generally agree on the timing of El Niño-related westerly wind anomalies, the magnitude of this interannual wind variability varies greatly among the products (Figure C-S1a).

Furthermore, biases in the historical wind speed data as a result of changes measurement methods through time complicate the analysis of zonal wind trends [e.g., Posmentier et al., 1989; Cardone et al., 1990]. As a result, these datasets also disagree regarding the zonal wind trend in the region surrounding Tarawa. Observations from Comprehensive

Ocean-Atmosphere Data Set (COADS) [Slutz et al., 1985] display a positive zonal wind trend, suggesting a weakening of the easterlies, while the historical gridded data and reanalysis products suggest a strengthening of the easterlies surrounding Tarawa (a negative trend).

We use quarterly and annually averaged zonal wind data from the 20th-century reanalysis project [Compo et al., 2011] to explore the relationship between zonal wind anomalies and Mn/Ca concentrations in the ~90-year long core from Tarawa atoll. We also compare to these records to δ18O from the same core and ENSO SST indices (Niño

3.4 and the El Niño Modoki (EMI) index [Ashok et al., 2007]) to further assess the connection between coral Mn/Ca and El Niño events. Annual averages of each series were calculated over the April-March period to preserve tropical climatology and ENSO events. We then compare trends in westerly winds and westerly bursts over the 20th century inferred from the Tarawa Mn/Ca record and the 6-hourly 20CR zonal wind data.

Westerly wind anomalies were defined as any positive 20CR zonal wind values, and westerly wind bursts were defined as westerly winds of greater than 5 m/s that lasted for at least 2 days [after Hartten, 1996]. Trends in both westerly wind anomalies and bursts

150 in the 6-hourly 20CR dataset were analyzed to address changes in both the short-lived wind events and persistent anomalies associated with El Niño events.

4. Results

4.1 A 90-year record of El Niño-related westerly winds

Mn/Ca in the Tarawa coral ranged from 8.3-80.2 nmol/mol between 1893 and

1982, with an average of 34.2 ± 10.2 nmol/mol. Consistent with Shen et al. [1992], we find that Mn/Ca is high during the major El Niño events of the 20th century (Figure C-

1b), when zonal winds in the region are anomalously westerly (positive zonal wind anomalies, Figure C-1c), and local conditions are warm and wet (negative δ18O anomalies, Figure C-1a). Quarterly and annual (April-March) Mn/Ca correlates

18 significantly with δ O (quarterly: rs=-0.33, N=357, P<0.001; annual: rs=-0.42, N=88,

P<0.001) and with ERSSTv3b Niño 3.4 SST anomalies (quarterly: rs=0.27, N=357,

P<0.001; annual:, rs=0.27, N=88, P=0.013), despite the fact that Mn/Ca tracks only the warm phase of ENSO. It also correlates with zonal wind anomalies from the 20CR project in the 10° by 10° gridbox surrounding Tarawa (quarterly: rs=0.24, N=357,

P<0.001; annual: rs=0.43, N=88, P<0.001). Quarterly Mn/Ca also displays a significant correlation with the EMI (quarterly: rs=0.15, N=88, P=0.005; annual: rs=0.12, N=88,

P=0.28), but the association is much weaker than with Niño 3.4 SSTs, suggesting that

Tarawa Mn/Ca primarily records westerly wind anomalies associated with classical eastern Pacific El Niño events. To test specifically for correspondence in the El Niño event signal among these records, we also performed a weighted rank correlation [Zar

151

1999; after Salama and Quade, 1982; Iman and Conover, 1987] to test for a correlation among the largest values of each timeseries. High annual Mn/Ca values correlate significantly with high annual zonal wind values (westerly wind anomalies, rT=0.43,

18 n=88, P<0.001) and with extreme (negative) annual δ O values (rT= -0.28, n=88,

P<0.01).

There is also a strong event by event correspondence in the strength of the Mn/Ca, zonal wind, δ18O and SST response during the major historical El Niño events defined by

Quinn et al., [1978, 1987] and by NOAA’s Oceanic Niño Index (ONI, 3 month running mean of ERSST.v3b SST anomalies in the Niño 3.4 region) (Table C-S1, Figure C-1).

To compare the strength of the El Niño signal among datasets, we calculated Z-scores

(normalized data where Z=(Xi-µ)/σ) for the Tarawa coral datasets, 20CR zonal wind, and the Nino3.4 and EMI SST indices using the mean and standard deviation of the period of overlap among records (1893.75-1982.75). We ranked each event as follows: (1) very weak events: Z=0.5-0.84 (~70-80th percentile), (2) weak events: Z=0.85-0.99 (~80-85%),

(3) moderate events: Z=1-1.2 (~85-90%), (4) strong events: Z=1.21-1.6 (~90-95%), and

(5) very strong events ≥ 1.6 (≥ ~95%), and compared the event strength to that originally defined by NOAA’s ONI (1950-present) and the Quinn et al. [1978, 1987] historical records (Quinn et al., [1978]: 1726-1976; Quinn et al., [1987]: 1525-1987). Although the historical El Niño events were generally stronger as defined here than in Quinn et al.,

[1978, 1987] or in the NOAA ONI, the Tarawa Mn/Ca record captured some of the strong historical El Niños that were underestimated or absent within the 20CR zonal wind dataset (1911-1912, 1917, 1932-1933, 1963-1964, 1968-1969 & 1976-1977). The 1917 and 1932-1933 events were also absent from the Niño 3.4 and EMI SST indices,

152 suggesting that wind anomalies in these years may not have triggered a basinwide event.

In contrast, only one weak to moderate El Niño event (1953-1954) was absent from the

Mn/Ca record. The Mn/Ca record also suggests El Niño conditions were observed at

Tarawa in 1946-1947, consistent with warm/wet conditions inferred from the δ18O record

[Cole et al., 1993]. These results provide strong support for the interpretation of Tarawa coral Mn/Ca as a proxy for El Niño-related westerly winds.

4.2 20th-century westerly wind trends

A reduction in the amplitude and frequency of Mn/Ca pulses occurs in the Tarawa record over the 1893-1982 period (Figure C-1b), suggesting a reduction in westerly wind anomalies. This change has produced a decrease in the overall Mn/Ca variance over the record (Figure C-2a), with highest Mn/Ca variability observed in the early 20th century.

The observed decrease in ENSO-related Mn/Ca variability over the 20th century, calculated as the ratio of the standard deviation of the high-pass filtered Mn/Ca series in the last half of the record to that of the first half of the record, is very similar to the observed decrease in ENSO-related variability in the network of available coral δ18O

th records over the 20 century [Thompson et al., 2011] (ratioMn/Ca(1900-1976)= 0.87; ratioδ18O(1895-1985)= 0.89; ratioδ18O(1900-1976)= 1.02). A similar mid-century reduction in variance was also observed in the zonal wind reanalysis, although zonal wind variance returns to early 20th century levels after 1980 (Figure C-2b). This increase in zonal wind variance in the late 20th century corresponds to frequent westerly wind bursts (westerly wind anomalies > 5m/s over 2 days) that occurred during the strong El Niño events of

1982-1983, 1991-1992 and 1997-1998 (red line in Figure C-2c, [Harrison and Vecchi

153

1997; Lengaigne et al. 2002]. The presence of high Mn/Ca variability during the early

20th-century, when westerly wind anomalies were frequent within the (6-hourly) 20CR dataset (black line in Figure C-2c), but burst events were rare (red line in Figure C-2c), suggests that Tarawa coral Mn/Ca is sensitive to persistent westerly wind anomalies, rather than to just strong short-lived burst events. The frequency of all westerly winds

(zonal wind > 0) displays a negative trend over the full 20th century (1871-2007, β=-

0.201, ns; Figure C-2c).

5 Discussion

Consistent with previous work, we find considerable disagreement among instrumental and reanalysis zonal wind data in terms of both the strength of interannual variability and the magnitude and direction of the trend in the region surrounding Tarawa

Atoll. Over the full period of coverage, COADs winds suggest a weakening of the easterly winds, while the historical gridded data and reanalysis products suggest a strengthening of the easterlies in this region. These discrepancies may reflect the limited number of observations, particularly prior to 1950, and they emphasize the need for an independent evaluation of tropical Pacific zonal wind trends.

Mn/Ca in corals from west-facing lagoons reflect westerly wind anomalies associated with El Niño events, and may therefore weigh in on tropical Pacific zonal wind trends. We demonstrate that Mn/Ca in a 90-year record from Tarawa atoll is significantly related to 20CR zonal wind anomalies, with strong positive Mn/Ca values and westerly wind anomalies during the major historical El Niño events. Additionally,

154 the coral Mn/Ca record from Tarawa atoll recorded events that were either absent or underestimated within the zonal wind reanalysis. These results provide strong support for the link between El Niño-related westerly wind anomalies and Tarawa coral Mn/Ca.

Manganese in corals may also respond to variations in atmospheric, volcanic or runoff-related input of Mn to the oceans, or variations in the strength of upwelling or advection of Mn-depleted waters [e.g., Shen and Boyle, 1988; Shen et al., 1991; Alibert et al. 2003; Carriquiry and Villaescusa, 2010]. However, these factors vary little at this small, remote equatorial atoll. For example, Shen et al. [1992] calculate that the atmospheric dust flux of Mn associated with the El Niño-related precipitation at Tarawa would only lead to a Mn anomaly of around 20%, compared to anomalies of over 50% in the coral Mn/Ca record. Further, this estimate was based on rainwater Mn concentrations from Bermuda [Jickells et al., 1984; Church et al., 1984] and is thus likely a conservative estimate, given that rainwater Mn concentration is typically orders of magnitude smaller at remote marine sites [e.g., Arimoto et al., 1985, 1987]. As deep waters near Tarawa are depleted in Mn relative to the surface [Shen et al., 1992; Slemons et al., 2010], a reduction in equatorial upwelling may contribute to anomalously high Mn/Ca during El

Niño events. Nonetheless, upwelling variations are small ( 1 C) west of the date line

[Wyrtki 1981] and would amplify the wind-driven Mn/Ca signal during El Niño events.

This 90-year Mn/Ca record suggests that there was a decrease in westerly wind anomalies in the western equatorial Pacific between 1893 and 1982. As westerly wind anomalies preferentially occur prior to and during El Niño events, when the tropical

Pacific warm pool is extended eastward [Yu et al., 2003; Vecchi and Harrison, 2000;

Eisenman et al., 2005], this is consistent with a reduction in El Niño-related variance

155 over this period. These results suggest a strengthening of the (predominantly easterly) zonal winds around Tarawa over the 1983-1982 period, as the frequency and impact of westerly wind anomalies are tied to the background state of the tropical Pacific and zonal wind strength [Fedorov 2002; Seiki and Takayabu 2007]. Zonal wind anomalies from the

20CR dataset also show a reduction in variance over this period; however, zonal wind variance returns to strong early 20th-century levels during the late 1980s and 1990s.

Nonetheless, a decrease in the frequency of westerly winds over the 20th-century (1900-

2000) was observed in the 20CR dataset (β= -1.2, N=100, P=0.038). Future studies are required to replicate this Mn/Ca record to capture the large events of the late 20th century and assess full 20th-century trends in westerly winds. Additional Mn/Ca records from westerly facing lagoons on other tropical Pacific atolls may be used to assess whether these trends reflect a change in the location of westerly wind bursts that precede El Niño events.

6 Conclusions and future work

The ~90-year Mn/Ca record from Tarawa atoll presented here supports earlier contentions that Mn/Ca in coral skeletons near west-facing lagoons provides a proxy for westerly winds. In the presence of large discrepancies among historical wind products, coral Mn/Ca records may provide key insights into variability and trends in the strength of zonal winds within the tropical Pacific. This 90-year Mn/Ca record from Tarawa atoll suggests that up to 1982, prior to the onset of that year’s strong El Niño, zonal wind variability weakened at this site. However, subsequent strong El Niño events (1982-1983

& 1997-1998) may have brought the zonal wind variance back to early 20th-century

156 levels. Nonetheless, the frequency of westerly winds in the 20CR dataset also shows a decline through the end of the 20th century, suggesting a potential strengthening of the mean zonal winds within the basin that warrants further investigation. Additional Mn/Ca records extending through the late 20th-century from islands with westerly facing lagoons are needed to replicate this proxy record and further assess 20th-century westerly wind trends. By adding Mn/Ca to the suite of coral tracers measured for paleoclimate reconstructions, we can expand our view of past ENSO variability to include westerly winds, along with the more commonly reconstructed variables of SST and salinity.

Acknowledgements

This research was supported by the NOAA climate program (award NA08OAR4310682),

The University of Arizona Department of Geosciences, and the Philanthropic Education

Organization.

Reference List

Alibert, C., et al. (2003). Source of trace element variability in Great Barrier Reef corals

affected by the Burdekin flood plumes. Geochimica et cosmochimica acta, 67(2),

231-246.

Arimoto, R., et al. (1985) Atmospheric trace elements at Enewetak Atoll: 2. Transport to

the ocean by wet and dry deposition. Journal of Geophysical Research 90, 2391-

2408.

157

Arimoto, R., et al. (1987) Trace elements in the atmosphere of American Samoa:

concentrations and deposition to the tropical South Pacific. Journal of Geophysical

Research 92, 8465-8479.

Ashok, K., et al. (2007). El Niño Modoki and its possible teleconnection. Journal of

Geophysical Research: Oceans (1978–2012), 112(C11).

Cardone, V. J., J. G. Greenwood, and M. A. Cane (1990) On trends in historical marine

wind data. J. Climate, 3, 113–127.

Carriquiry, J. D. and J.A. Villaescusa (2010) Coral Cd/Ca and Mn/Ca records of ENSO

variability in the Gulf of California, Clim. Past, 6, 401-410, doi:10.5194/cp-6-401-

2010.

Chen, J., B.E. Carlson, and A.D Del Genio (2002), Evidence for strengthening of the

tropical general circulation in the 1990s. Science, 295(5556), 838-841.

Church, T. M., et al. (1984) The wet deposition of trace metals to the western Atlantic

ocean at the mid-Atlantic coast and on Bermuda, Atmos. Environ., 18, 2,657–2,664

Cole, J. E., R.G. Fairbanks, and G.T. Shen (1993). Recent variability in the Southern

Oscillation: Isotopic results from a Tarawa Atoll coral. Science, 260(5115), 1790-

1793.

Cole, J. E., and R.G. Fairbanks (1990). The Southern Oscillation recorded in the δ18O of

corals from Tarawa Atoll. Paleoceanography, 5(5), 669-683.

Compo, G. P., et al. (2011), The Twentieth Century Reanalysis Project. Q.J.R. Meteorol.

Soc., 137: 1–28. doi: 10.1002/qj.776.

158

Deser, C., A.S. Phillips, and M.A. Alexander (2010), Twentieth century tropical sea

surface temperature trends revisited. Geophysical Research Letters, 37(10).

Eisenman, I., L. Yu, and E. Tziperman (2005) Westerly Wind Bursts: ENSO’s Tail

Rather than the Dog?. J. Climate, 18, 5224–5238.

Fedorov, A.V. (2002) The response of the coupled tropical ocean-atmosphere to westerly

wind bursts. Q. J. Roy. Meteorol. Soc. 128, 1-23.

Feng, M., M.J. McPhaden, and T. Lee (2010), Decadal variability of the Pacific

subtropical cells and their influence on the southeast Indian Ocean. Geophysical

Research Letters, 37(9), L09606.

Feng, M., et al. (2011), The reversal of the multi‐decadal trends of the equatorial Pacific

easterly winds, and the Indonesian Throughflow and Leeuwin Current transports.

Geophysical Research Letters, 38(11).

Harrison, D. E., G.A. Vecchi (1997) Westerly Wind Events in the Tropical Pacific, 1986–

95*. J. Climate, 10, 3131–3156.

Hartten, L. M. (1996). Synoptic settings of westerly wind bursts. Journal of geophysical

research, 101(D12), 16997-17.

Iman, R. L., and W.J. Conover (1987). A measure of top–down correlation.

Technometrics, 29(3), 351-357.

Jickells, T. D., A.H. Knap, and T.M. Church (1984). Trace metals in Bermuda

159

rainwater. Journal of Geophysical Research: Atmospheres, 89(D1), 1423-

1428.

Lengaigne, M., et al. (2002). Ocean response to the March 1997 westerly wind event.

Journal of Geophysical Research: Oceans, 107(C12), SRF-16.

Lengaigne, M., et al. (2003). The March 1997 westerly wind event and the onset of the

1997/98 El Nino: Understanding the role of the atmospheric response. Journal of

climate, 16(20), 3330-3343.

Li, G., and B. Ren (2012), Evidence for strengthening of the tropical Pacific Ocean

surface wind speed during 1979–2001. Theoretical and Applied Climatology, 107(1-

2), 59-72.

Meehl, G.A., et al. (2007), Global climate projections, in Climate Change 2007: the

physical science basis. Contribution of Working Group I to the Fourth Assessment

Report of the Intergovernmental Panel on Climate Change, edited by S. Solomon, et

al., pp. 747–845, Cambridge University Press, Cambridge.

Merrifield, M. A. (2011). A shift in western tropical Pacific sea level trends during the

1990s. Journal of Climate, 24(15), 4126-4138.

Merrifield, M. A., and M.E. Maltrud (2011), Regional sea level trends due to a Pacific

trade wind intensification. Geophysical Research Letters, 38(21).

Posmentier, E. S., M. A. Cane, and S. E. Zebiak (1989) Tropical Pacific climate trends

since 1960. J. Climate, 2, 731–736.

160

Power, S. B., and G. Kociuba (2011), What caused the observed twentieth-century

weakening of the Walker circulation?. Journal of Climate, 24(24), 6501-6514.

Power, S. B., and I.N. Smith (2007), Weakening of the Walker Circulation and apparent

dominance of El Niño both reach record levels, but has ENSO really

changed?. Geophysical Research Letters, 34(18).

Quinn, W. H., et al. (1978), Historical trends and statistics of the Southern Oscillation, El

Niño, and Indonesian droughts. Fish. Bull, 76, 663-678.

Quinn, W. H., V.T. Neal, and S.E.A. De Mayolo (1987), El Niño occurrences over the

past four and a half centuries, J. Geophys. Res., 92(C13), 14449–14461,

doi:10.1029/JC092iC13p14449.

Salama, I., and D. Quade (1982), A nonparametric comparison of two multiple

regressions by means of a weighted measure of correlation. Communications in

Statistics-Theory and Methods, 11(11), 1185-1195.

Seiki, A., and Y.N. Takayabu (2007), Westerly wind bursts and their relationship with

intraseasonal variations and ENSO. Part I: Statistics. Monthly Weather

Review, 135(10), 3325-3345.

Shen, G. T., and E.A. Boyle (1988). Determination of lead, cadmium and other trace

metals in annually-banded corals. Chemical Geology, 67(1), 47-62.

Shen, G. T., et al. (1991), Paleochemistry of manganese in corals from the Galapagos

Islands. Coral Reefs, 10(2), 91-100.

161

Shen, G.T., et al. (1992), A chemical indicator of trade wind reversal in corals from the

western tropical Pacific, Journal of Geophysical Research, 97 (C8), 12698-12697,

doi: 10.1029/92JC00951.

Slutz, R. J., et al. (1985), Comprehensive ocean-atmosphere data set: Release 1. I:

Climate Research Program, ERL/NOAA, Boulder, CO.

Smith, T.M., et al. (2008) Improvements to NOAA's Historical Merged Land-Ocean

Surface Temperature Analysis (1880-2006), Journal of Climate, 21, 2283-2296.

Thompson, D. M., et al. (2011), Comparison of observed and simulated tropical climate

trends using a forward model of coral δ18O. Geophysical Research Letters, 38(14).

Tokinaga, H., et al. (2012), Regional Patterns of Tropical Indo-Pacific Climate Change:

Evidence of the Walker Circulation Weakening*. Journal of Climate,25(5), 1689-

1710.

Vecchi, G. A., and D.E. Harrison (2000), Tropical Pacific Sea surface temperature

anomalies, El Niño, and Equatorial Westerly Wind Events*. Journal of

climate, 13(11), 1814-1830.

Vecchi, G. A., et al. (2006), Weakening of tropical Pacific atmospheric circulation due to

anthropogenic forcing. Nature, 441(7089), 73-76.

Vecchi, G. A., and B.J. Soden (2007), Global warming and the weakening of the tropical

circulation. Journal of Climate, 20(17), 4316-4340.

162

Wu, R., and S.P. Xie (2003), On Equatorial Pacific Surface Wind Changes around 1977:

NCEP-NCAR Reanalysis versus COADS Observations*. Journal of climate, 16(1),

167-173.

Wyrtki, K. (1981), An Estimate of Equatorial Upwelling in the Pacific. J. Phys.

Oceanogr., 11, 1205–1214. doi: http://dx.doi.org/10.1175/1520-

0485(1981)011<1205:AEOEUI>2.0.CO;2

Yu, L., R.A. Weller, and W.T. Liu (2003), Case analysis of a role of ENSO in regulating

the generation of westerly wind bursts in the western equatorial Pacific. Journal of

Geophysical Research: Oceans (1978–2012), 108(C4).

Zar, J. H. (1999). Biostatistical analysis. 4th. Ed. Prentice Hall. USA.

163

Figure C-1: (inset) Google Earth image of Tarawa atoll, Republic of Kiribati (1.4°N, 173°E) showing the location of coral core collection (star) and the westward facing lagoon with Mn enriched sediments. Westerly winds (black arrows), associated with El Niño events, remobilize Mn from these sediments. Mn enriched seawater may then be transported to the nearby fringing reef and incorporated into the coral skeleton (white arrow). Quarterly (a) δ18O and (b) Mn/Ca records from the Tarawa atoll coral and (c) Zonal wind (m/s) from the 20th Century Reanalysis project (20CR) in a 10 x 10 degree gridbox surrounding Tarawa atoll [Compo et al., 2011], where red shading denotes historical El Niño events [Quinn et al., 1978; Quinn et al., 1987; NOAA Oceanic Niño Index (ONI)])

Figure C-2: (a) Mn/Ca variance in 20-year moving windows, (b) 20CR zonal wind variance in 20-year moving windows, and (c) 20CR westerly winds (6-hourly 20CR zonal wind >0, gray), 20CR annual westerly wind frequency (black), and 20CR annual westerly wind burst frequency (red, westerly winds > 5 m/s lasting for 2 days [after Hartten, 1996]).

164

Figure C-S1: (a) Annual zonal wind anomalies (m/s) in the 10 x 10 degree region (4S-6N, 166-176E) surrounding Tarawa atoll, and (b) number of COADS observations per month in this 10° x 10° region (note: number of observations peaked at 913 in February 1944). Zonal wind anomalies were calculated relative to the full period of coverage in each wind product: 20th-century reanalysis (“20CR”) [Compo et al., 2011], NCEP reanalysis [Kalnay et al., 1996], Comprehensive Ocean-Atmosphere Data Set (COADs) [Slutz et al., 1985], Weare marine climate atlas [Weare et al., 1980], and Atlas of Surface Marine Data 1994 [da Silva et al., 1994].

Supplemental references da Silva, A. M., C.C. Young, and S. Levitus (1994). Atlas of surface marine data 1994,

Vol. 4: Anomalies of fresh water fluxes. Noaa atlas nesdis, 9, 308.

Kalnay, E., et al. (1996) The NCEP/NCAR 40-Year Reanalysis Project. Bulletin of the

American Meteorological Society, 77(3), 437-471.

Weare, B. C., P.T. Strub, and M.D. Samuel (1980), Marine climate atlas of the tropical

Pacific Ocean. Department of Land, Air and Water Resources, University of

California.

165

Table C-S1: Strength of historical El Niño events defined by Quinn [(1) 1978, (2) 1987] and by (3) NOAA’s Oceanic Niño Index (ONI) (left column), compared to strength defined from Z-scores of Tarawa coral Mn/Ca, δ18O, 20CR winds, Nino3.4 SST, and the El Niño Modoki index (see text). To account for the sign of the El Niño signal in each record, the maximum Z scores were used for Mn/Ca, wind and SST indices, and minimum Z scores were used for δ18O. If the strength is not indicated, then the event was not observed in that dataset (although note that the ONI only covers from 1950-present). VW, W, M, S, and VS denote very weak, weak, moderate, strong and very strong events, respectively.

Event Mn/Ca δ18O 20CR wind Niño 3.4 SSTa EMI

1896-1897 (M1) S VS VS VS M

1899-1900 (S1,2) VS VS VS VS S

1902-1903 (M1) VS S VS VS VS

1905-1906 (M1) M VS VS VS S

1911-1912 (S1,2) VS VS M VS VW

1914-1915 (M1) VS VS VS S S

1917 (S2) VS

1918-1919 (M-S1) VS S S S W

1923-1924 (W1) S S VS M VS

1925-1926 (S1-VS2) VS VS S VS VW

1930-1931 (M1) VS S VS S VS

1932-1933 (W1,S2) VS VW

1939-1940 (W-M1) M VS VS S M

1940-1941 (S2) VS VS VS VS VS

1941-1942 (S1) VS VS S VS VS

1951-1952 (W1,M3) VW M VW S M

1953-1954 (M1,W3) M S M VS

1957-1958 (S1,2,3) VS VS VS VS VS

1958-1959 (W3) VS VS VS VS VS

1963-1964 (W3) VW S S M

166

1965-1966 (M1, S3) VS VS VS VS VS

1968-1969 (M3) S M W VS VS

1969-1970 (W1,3) VW M W VS VS

1972-1973 (S1,2,3) VS VS VS VS VW

1976-1977 (M1,W3) VS VS VW S VS

1977-1978 (W3) VW S VS S VS

167

APPENDIX D

CLIMATE CONTROLS ON BAINBRIDGE CRATER LAKE, GALÁPAGOS OVER THE PAST 6000 YEARS

To be submitted to the professional journal: Journal of Paleolimnology

168

CLIMATE CONTROLS ON BAINBRIDGE CRATER LAKE, GALÁPAGOS OVER THE PAST 6000 YEARS

Diane M. Thompson, Jessica L. Conroy, Aaron Collins, Jonathan T. Overpeck, Julia E. Cole, and Mark B. Bush

Abstract

It has been proposed that the finely laminated sediments within Bainbridge Crater Lake,

Galápagos provide a record of El Niño-Southern Oscillation (ENSO) events over the

Holocene. However, current hypotheses for how the climate signal gets preserved in the lake sediment record have not been directly tested. In this work, we monitored the local climate and limnology of Bainbridge Crater Lake to determine how seasonal to interannual climate variability translates into the lake sediment record. Monitoring results suggest that the brown-green, organic rich, siliciclastic laminae in the Bainbridge sediment record indicate warm, wet conditions typical of El Niño events, when terrigenous material is washed into the lake from the crater walls and dissolved inorganic carbonate (DIC) concentration remains low. In contrast, lake monitoring suggests that carbonate laminae form when DIC is concentrated in the lake water as a result of cool, dry conditions typical of La Niña events. Applying these findings to the climate interpretation of the Bainbridge Crater Lake sediment record, we find ENSO events of both phases were reduced during the mid-Holocene (~6100-4000 cal. years BP) relative to the last ~2500 cal. years. Abundant carbonate laminations were observed between

4000 and 3500 cal. years BP, when conditions in the Galápagos region were cool and dry and the tropical Pacific SST gradient may have strengthened. Following reduced mid-

Holocene ENSO, the Bainbridge sediment record suggests a stepwise increase in ENSO

169 variability starting around 2500 cal. years BP. Strong interannual variability persisted from 1750 to ~900 cal. years BP, consistent with a reduced tropical Pacific zonal SST gradient and increased frequency of ENSO events in other regional records. When considered together, the Bainbridge sediment record and all other available ENSO reconstructions from the tropical Pacific suggest that a marked increase in ENSO-related variability occurred around 1775 (± 190) cal. years BP. After ~900 years BP, sedimentation rate at Bainbridge slowed dramatically and overall ENSO-related variability at the lake decreased as the tropical Pacific zonal SST gradient strengthened.

Over the past 900 years, intervals of gypsum precipitation suggests that La Niña-like conditions dominated between 400-600 BP and between ~350 and 0 cal. years BP (AD

1950), whereas wetter, El Niño-like conditions dominated in the most recent 60 years.

Introduction

Global climate patterns are strongly impacted by interannual and low-frequency changes in the tropical Pacific Ocean sea-surface temperatures (SSTs), particularly in response to departures associated with El Niño-Southern Oscillation (ENSO) events.

However, despite the importance of the tropical Pacific to climate on a global scale, observational and model products disagree regarding recent temperature trends in the basin (e.g., Vecchi et al. 2008; Deser et al. 2010; Thompson et al. 2011; Yeh et al. 2012), and global climate models differ widely in their projections of tropical Pacific mean state and ENSO variance under future scenarios of increased greenhouse gas concentrations

(e.g., Meehl et al. 2007; Guilyardi et al., 2009; Collins et al. 2010; Guilyardi et al. 2012;

170

Kim and Yu 2012; Stevenson et al. 2012). The discrepancy between observational temperature records is particularly apparent in the eastern Pacific, where instrumental data products indicate a warming trend or slight cooling trend over the 20th century

(Vecchi et al., 2008; Deser et al. 2010), whereas in situ SST measurements off of Isla

Santa Cruz, Galápagos archipelago show no significant trend (Trueman and d’Ozouville

2010; Wolff 2010). The sign of the trend within these datasets may also differ by season

(Karnauskas et al. 2009; Wolff 2010) and time period (Liu et al. 2005). Understanding recent climate variability in the eastern Pacific is critical not only because this region is strongly affected by ENSO anomalies, but also because upwelling trends in this region may be key to understanding the ocean’s response to future radiative forcing (Clement et al. 1996; Karnauskas et al. 2009).

Paleoclimate records from the tropical Pacific may help resolve the discrepancy regarding recent trends within the basin and place these changes into the context of variability observed over the past few millennia or longer. Climate reconstructions from corals and marine and lake sediments have already provided key insights into changes in ocean temperature, precipitation and ENSO variance in the eastern tropical Pacific. For example, paleoclimate reconstructions provide growing evidence for an intensification of

ENSO variability in the late Holocene (Moy et al. 2002; Riedinger et al. 2002; Koutavas et al. 2006; Conroy et al. 2008; Donders et al. 2008). However, there are still a limited number of paleoclimate records from the tropical Pacific with high enough temporal resolution to separate interannual to decadal variability from changes in background state of the basin, yet are also long enough to study century-scale climate trends. High- resolution lake sediment records may provide such constraints on seasonal to millennial

171 tropical climate variability, but available records display considerable disagreement regarding multidecadal to centennial climate changes over the past millennium. These discrepancies may be attributed at least in part to uncertainties surrounding the climatic interpretation of the records, stemming from a lack of modern observations with which to calibrate the lake sediment variables. Although such modern calibrations are critical to understanding how the climate signal gets preserved in the sediment records, there have been no published analyses to date that characterize limnological variability over the course of years or longer in tropical Pacific island lakes due to the difficulty in reaching and revisiting these remote sites.

Sediment records from Bainbridge Crater Lake, Galápagos archipelago (0°21'S,

90° 34'W) have great potential as proxy records of past changes in tropical Pacific climate variability. For example, laminations preserved within the Bainbridge stratigraphy have been interpreted as a record of the frequency of moderate and strong El

Niño events during the late Holocene (Riedinger et al. 2002). Riedinger et al. (2002) proposed that rainfall (and/or influx of seawater) associated with El Niño events of different intensities led to the deposition of distinct carbonate and siliciclastic laminae that were preserved in the lake sediment record. They proposed that siliciclastic laminae form as a result of erosion of the crater walls during the intense rainfall characteristic of strong events, and carbonate laminae form during events of moderate intensity, when rainfall is strong enough to promote lake stratification and carbonate precipitation, but not significant erosion of the crater walls. In 2009, we began monitoring the local climate and limnology of Bainbridge Crater Lake to explore these hypotheses for sedimentation in the lake and to determine how seasonal to interannual climate variability

172 is recorded in the Bainbridge lake sediment record. In this work, we present the results of monitoring the climate and limnology of Bainbridge Crater Lake from December 2009 to

October 2012. Based on these monitoring results, we propose a new and distinctly different hypothesis for laminae formation within the Bainbridge sediment record.

Applying this new climate interpretation to the sediment record, we then discuss the implications for tropical climate variability over the past few thousand years.

Bainbridge Crater Lake

Bainbridge Crater Lake lies at sea level in the largest of the Rocas Bainbridge volcanic islands along the southeastern coast of Isla Santiago, Galápagos. The lake is small and shallow (~0.2 km in diameter and < 3.3 meters deep) and is nearly surrounded by moderately steep crater walls (Figure D-1). The crater walls taper along the south side of the island to a low point that sits ~2-3 meters above sea level. The lake is hypersaline, with an average salinity about 3 times that of seawater. The hypersaline nature of the lake suggests that the lake is likely connected to the ocean through fissures in the basalt and through sea spray and storm surge over the lip of the crater wall during large storm events. The vegetation on the island is dominated by the salt-tolerant Sesuvium portulacastrum and Batis maritima plants, with Croton sp. bushes lining the crater rim.

American flamingos (Phoenicopterus ruber) wade and forage in the shallow northern shore of the lake, but the sediments in the deeper sections of the lake (sampled here) remain out of reach and thus undisturbed.

173

Materials and Methods

Environmental monitoring

In December 2009, we deployed a weather station and lake sondes to monitor the environmental conditions in and around the lake. HOBO weather stations recorded precipitation, air temperature, solar radiation, wind speed and wind direction at 15 minute intervals from December 17, 2009 to June 6, 2010 and at 30-minute intervals from June

6, 2010 to October 3, 2012. To monitor the effect of local climate variability on lake conditions, we also deployed AquaTroll 100 sondes to measure temperature, conductivity, salinity, total dissolved solids (TDS) and density over the same sampling intervals. Sondes were deployed at 1 and 2 meters depth near the deepest section of the lake (~3.1 meters depth). A GPS sounding device was used to identify the deepest section of the lake and to create a bathymetric map of the lake (Figure D-1c). All of the instruments were cleaned, maintained and re-launched on subsequent trips in June 2010 and October 2012. Daily average meteorological and limnological conditions were calculated from the weather station and sonde measurements and a large sample

Wilcoxon Signed Ranks test (Z statistic) was utilized to assess the median differences between the sampling periods (paired, nonparametric data). As a result, we have nearly continuous daily-resolved data for the climate and limnology of Bainbridge Crater Lake at daily resolution from December 2009 to October 2012, a period that covers more than two full seasonal cycles and includes both a moderate El Niño (2009-2010 warm season) and a moderate La Niña (the 2010-2011 warm season). Daily averages of the local

174 meteorological data from the weather station were compared to data collected from the

Charles Darwin Research Station in Puerto Ayora, Isla Santa Cruz (Charles Darwin

Foundation 2012) and from the Tropical Atmosphere Ocean (TAO) moored buoy at

0ºN95ºW (Figure D-1). We used the energy budget equation (Shuttleworth 1993) to calculate evaporation at the lake surface in mm/day from local air temperature, lake temperature, solar radiation, and wind speed along with relative humidity from Puerto

Ayora (Charles Darwin Foundation 2012).

In addition to the long-term monitoring of the lake, we collected profiles through the water column in December 2009, June 2010 and October 2012. A handheld YSI 85 was used to collect profiles of temperature, dissolved oxygen (DO), salinity and conductivity and a van Dorn sampler was used to collect water samples every meter throughout the water column in the deepest portion of the lake. The ratio of the stable oxygen isotopes (δ18O) in these water samples was determined using the Thermo Delta

XP plus, Gas Bench II, Isotope Ratio Mass Spectrometer at the University of Arizona.

The ratio of deuterium to hydrogen (δD) was also determined on a gas-source isotope ratio mass spectrometer (Finnigan Delta S) in the Environmental Isotope Laboratory at the University of Arizona. Analysis of the major and minor cations and anions, alkalinity and total inorganic carbon (TIC: dissolved CO2, carbonates, and bicarbonates) were performed in the Department of Hydrology and Water Resources. The concentration of major and minor cations and major and minor anions were measured using an ICP-OES and ion chromatography, respectively, and alkalinity was measured by acid titration.

175

Sediment sampling

We collected additional short and long cores from the lake to add to previous analyses of the sediment record from cores collected in December 1991 (Riedinger et al.

2002). Three cores were collected in 2007 by the Florida Institute of Technology paleoecology group using a Colinvaux-Vohnout piston corer, and four additional short

(~77 cm to ~99 cm long) sediment cores were taken in December 2009 by the University of Arizona Geosciences group using an Aquatic Research Instruments Universal

Percussion Corer with a 1 meter long sampling tube. The sediment cores were collected from the deepest section of the lake in approximately 3-3.2 meters depth. The mud- water interface was preserved in each core, and we extruded the uppermost sediments in the field due to their high water content. The sediment cores were brought back to the laboratory for analysis, where they were subsequently split, photographed, and refrigerated. Further analyses focused on the 3.6m piston core collected in 2007 and a

92cm universal core collected in 2009 that displayed the best laminations.

To understand how the local environmental and limnologic conditions were recorded in the lake sediments, sediment traps were deployed to collect sediments settling to the lake bottom. The sediment traps were deployed at 1 and 2 meters depth along the same line as the sondes (in ~3.1 meters depth) on December 2, 2009 and June 6, 2012.

Sediment trap samples covering the two collection intervals (12/2/2009-6/5/2010 and

6/6/2010-10/2/2012) were rinsed eight times with milli-Q water to remove salts and freeze dried prior to further analysis. Ekman dredge sampling throughout the lake was also used to obtain a snapshot of the surface sediments prior to the onset of the El Niño event in December 2009. Finally, to improve our interpretation of the chemical

176 variations in the sediment cores, samples of rock varying with degrees of visible weathering and a soil sample were collected from locations around the rim of the lake.

The soil sample was ground with a mortar and pestle and a 1:1 ratio of the sample and distilled water was stirred for half hour for estimation of soil pH (Buol et al. 1997).

We used a CAMECA SX50 electron microprobe in the Department of Planetary

Sciences at the University of Arizona to map the distribution of Ca, Sr, Si, Fe, Ti, K, S,

Mg, P, Mn, and Na in sediment trap samples from 2 meters depth as well as the parent rock material. Samples were mounted on carbon tape and carbon coated for analysis.

Four wavelength dispersive spectrometers (WDS) were run simultaneously (TAP: Si,

Mg, Na, TAP: Sr, Al, P, PET: Ca, K, S, and LIF: Fe, Ti, Mn), with an accelerating voltage of 20 kV, a beam current of 40 nA, a beam size of 16 μm, and a pixel time of 8 ms. Back-scattered electron (BSE) images and Energy Dispersive Spectroscopy (EDS) were then utilized to investigate the crystallography and chemical composition of each sample component identified from the maps.

In addition to microprobe analysis, an EDAX Eagle III tabletop scanning μ-XRF analyzer in the Geosciences Department at the University of Arizona was used to determine the concentration of major and trace elements in rock and sediment trap samples. The μ-XRF has a Mo target and was set to generate incident X-rays with an energy of 40 KV and 400 mA. Measurements of major and trace element intensities (Na,

Mg, Al, Si, P, S, Cl, K, Ca, Ti, V, Mn, Fe, Cu, Zn, As, Sr, Mo, and Tc) were made over a

15 second integration time at 100 μm spacing along 2 line scans spanning pellets of compressed sample material. Elemental intensities were also measured down the sediment core at 100μm increments. Sediment core samples were embedded in epoxy

177 resin, and an average of seven parallel linescans, spaced 1 mm apart, was used to optimize the signal to noise ratio due to the heterogeneity of the elements within the sediment core. Normalized ratios with respect to potassium were used in all subsequent analyses to remove variability associated with changes in the thickness of the embedded sediments throughout the core. Two gaps in the XRF record occur in the intervals containing unconsolidated gypsum gravel that could not be embedded in epoxy resin.

A GEOTEK core logger at the University of Florida LUCIE group generated high-resolution photographs of the three 2007 cores for grayscale analysis

(http://www.geotek.co.uk/services/core_logging). Due to sediment slumping that occurred within the deepest portion of the lake, we created a composite core from the images of the two longest cores to obtain a complete record free of any slumping effects.

A record of rapid sediment color changes was then produced from this composite core using Image J v. 1.0 grayscale analysis (Rasband 1999).

Samples for pollen analysis were taken from the longest (~3.6 meter) 2007 core during periods of color and sediment composition transitions (n=30, 0-170 cm). Samples were treated using standard pollen preparation protocols (Faegri and Iversen 1989). To calculate pollen concentration (Battarbee and Kneen 1982), each sample was spiked with either 5000 or 10,000 polysterene microspheres based on the volume of the sample

(either 0.5 cm3 or 1 cm3, respectively). Pollen counts were conducted until 100 pollen grains or 2000 microspheres were counted. Due to low pollen concentrations within the core, total numbers of pollen grains were used in further analyses. Total pollen grain counts were also split into local (Sesuvium sp. and Croton sp.) and regional (other)

178 sources to separate the local rainfall-driven vegetation response from the regional wind- driven pollen signal.

Finally, these new records from Bainbridge Crater Lake were compared with the original laminae record from this lake (Riedinger et al. 2002) and reconstructions of

ENSO variability (Moy et al. 2002, Rein et al. 2005, Conroy et al. 2008, Makou et al.

2010) and tropical Pacific SSTs (Koutavas et al. 2002, Stott et al. 2004, Rein et al. 2005,

Oppo et al. 2009) to assess implications for tropical Pacific climate variability over the past 6000 years. For this comparison, we counted laminae in the Riedinger et al. (2002) sediment record in 250 and 500-year bins to assess millennial-scale variability at the lake.

Age modeling

Due to the absence of microfaunal or floral material within the cores, we established an age model for the sediment record through radiocarbon dating of bulk sediment samples. Building off of the original chronology for the lake based on dates from 49 bulk sediment samples (Riedinger et al. 2002), we generated an additional 10 dates from a short core collected in 2009 and 4 dates from a long core collected in 2007.

Surface algal samples collected at the sediment-water interface by the Ekman dredge in

December 2009 and dissolved inorganic carbon (DIC) from water samples collected at 3 meters depth in December 2009 and October 2012 were used to determine the reservoir age of the lake. The standard HCl-NaOH-HCl procedure was used to pretreat the algal samples, while an acid-only treatment was used for the bulk sediment samples. Bulk sediment samples from the 2007 core (n=4) were processed at the National Ocean

179

Sciences Accelerator Mass Spectrometry Facility (NOSAMS), whereas the rest of the samples were combusted and measured for radiocarbon content at the University of

Arizona Accelerator Mass Spectrometry Laboratory. The reservoir age of the lake determined in this work was used along with the radiocarbon dates from the 49 bulk sediment samples (Riedinger et al. 2002) to modify the original chronology for the 1991 sediment core (Riedinger et al. 2002).

210Pb and 137Cs abundance in the extruded surface sediments was also measured by the Land Use and Environmental Change Institute at the University of Florida.

However, a 210Pb age model could not be obtained for the lake, as no measurable Pb-210,

Ra-226 or Cs-137 activity was detected in these uppermost extruded surface sediments, despite our high confidence that we collected an intact sediment-water interface. A similar absence of Pb-210 activity in surface sediments has been noted in other hypersaline Galápagos lakes (M. Brenner and W. Kenney, personal communication).

Calibration of the radiocarbon dates and age-depth modeling utilized the Clam package v2.1 in R (Blauuw 2010). We used a mixed calibration curve based on the reservoir age of 87 14C years ±35 (Table D-3) determined from the DIC of the 2012 water sample from 3 meters depth. Calibrated 14C ages and 2σ weighted average standard deviations were reported in years BP (with respect to AD 1950). A spline with a smoothing weight of 0.3 was used for age-depth modeling of the 2007 and 2009 cores, and a second-order polynomial was used for the age-depth modeling of the 1991 core

(following the approach of Riedinger et al. 2002). The median and 95% confidence intervals of iterations without age reversals were used to develop the chronology of the cores from 1991 (1,000 iterations, no reversals), 2007 (100,000 iterations, no reversals)

180 and 2009 (1899 iterations without reversals). Distinctive laminations and marker bands were used to correlate laminae between the cores.

Results

The results section is divided as follows. First, we summarize the results of monitoring the local climate and limnology at Bainbridge Crater Lake between December

2009 and October 2012, emphasizing factors likely to drive differences in the composition of sediments deposited in the lake between the two sampling intervals

(section 5.1). We then compare the composition of the sediments collected during the two sampling periods (December 2009-June 2010 and June 2010-October 2012) and compare the composition of the sediments with that of the parent rock material and regional atmospheric dust, which may serve as exogenous sources of sediment to the lake

(section 5.2). In section 5.3 we discuss the age model for the Bainbridge sediment record and its implications for sedimentation rate in the lake. Finally, in section 5.4, we discuss the implications of the Bainbridge sediment record for tropical Pacific climate over the past six thousand years.

Climate and limnology

Bainbridge Crater Lake climate mirrors that of the Galápagos more generally, and has two distinct seasons: a warm/wet season from January through May, and a cool/dry season from June through December (Mitchell and Wallace 1992; Trueman and

181 d’Ozouville 2010; Figure D-2, Figure D-3). Daily average air temperature reached 26-

28ºC in the warm season and fell to as low as 19ºC in the cool season during our monitoring period. Solar radiation and calculated evaporation were greater during the warm season, with a daily average of ~450 W/m2 of radiation incident on the lake and

~11 mm/day of evaporation during the warm season, compared to ~420 W/m2 and 9.5 mm/day during the cool season. Stronger, more southeasterly winds were observed in the cool season, with daily average wind speeds of ~4 m/s and gusts of up to 10 m/s.

Precipitation at the lake fell almost exclusively in the warm season when daily air temperatures rose above 24ºC, with an average of ~1 mm/day and a maximum of 80 mm/day during the moderate 2010 El Niño event. Although no precipitation was recorded at Bainbridge during the 2012 warm season, an average of ~2.9 mm/day was recorded by the station at Puerto Ayora during this period (see Figure D-1). As precipitation at Bainbridge and Puerto Ayora was correlated during the initial logging period (12/18/2009-6/5/2010; rs=0.49, N=170, P<0.001), the absence of precipitation during the 2012 warm period may be an artifact of a malfunction of the Bainbridge rain gauge. We therefore include Puerto Ayora station data in all further precipitation analyses.

Bainbridge Crater Lake is shallow and well mixed by strong SE trade winds blowing across the lake surface from a low point on the south side of the crater wall (see

Figure D-1b). The lake therefore displays strong seasonality in response to the local climatology (Figure D-2, 4), with significant differences (warm-cool) in temperature, salinity, density and TDS between seasons (temperature: Z=16.8, N=1037, P<0.001; salinity: Z=-4.995, N=1037, P<0.001; density: Z=-9.53, N=1037, P<0.001; TDS: Z=-

182

5.97, N=1037, P<0.001). YSI profiles of temperature, conductivity and dissolved oxygen collected from the deepest portion of the lake on the afternoon of June 6, 2010 (Table D-

1) and the close correspondence of temperature and salinity at 1 and 2 meters depth throughout the monitoring interval (Figure D-2) are consistent with a well-mixed lake.

Daily average water temperatures were around 32ºC during the warm season (with a maximum of 42ºC), whereas temperatures fell to around 25ºC during the cool season. In response to changes in the local P-E balance, the lake also became more dense during the cool season, with higher TDS. The daily average salinity rose from 99 PSU during the warm season to 102 PSU during the cool season, and the average density rose from 1.07 g/cm3 to 1.075 g/cm3, although these differences were not significant.

Interannual variability associated with ENSO events was also observed during the warm season at Bainbridge Crater Lake. The warm season of 2010, a moderate El Niño year, was significantly warmer and wetter than the warm season of 2011, a moderate La

Niña (Figure D-5; 2010-2011 temperature: Z=3.511, N=151, P<0.001; 2010-2011 precipitation: Z=4.278, N=151, P<0.001). Daily average evaporation at the lake was also significantly lower during the 2010 warm season than in both the 2011 and 2012 warm seasons (2010-2011: Z= -2.387, df=151, P=0.017; 2010-2012: Z=-2.402, df=151,

P=0.016). As a result, daily average salinity, density and TDS were all higher in the

2011 warm season than observed in the 2010 warm season, although these differences were not significant. The cooler, denser lake conditions are consistent with a negative

P/E budget and cool, dry La Niña conditions in the 2011 warm season. However, the lowest daily salinity, density and TDS values were actually observed during the 2012 warm season despite relatively low precipitation (at both Bainbridge and Puerto Ayora)

183 and high evaporation. Specific electrical conductance of the water samples collected in

December 2009, June 2010 and October 2012 are in good agreement with the specific conductivity measured by the sonde (Figure D-S1), suggesting that these trends in salinity cannot be attributed to instrumental drift. Instead, the low salinity and TDS in the lake in 2012 may be attributed to increased mineral precipitation during this dry period, pulling cations and anions out of solution an into the solid phase (see further discussion of water chemistry below).

As a result of this seasonal to interannual climate variability, the two sediment trap sampling periods (Figure D-2) were characterized by distinctly different climate and limnology. The first sampling period (December 2, 2009- June 5, 2010) covered one warm season and moderate El Niño event, whereas the second sampling period (June 6,

2010- October 2, 2012) covered two cool seasons and two warm seasons, with one warm season being a moderate La Niña (Figure D-2). The local climate and limnology was therefore more variable during the 2010-2012 (“2012”) sediment trap period than in the

Dec-June 2010 (“2010”) sediment trap period (Figures D-7, D-8). Nonetheless, the 2010 sediment trap period was significantly warmer (2010-2012: Z=12.62, N=1020, P<0.001) and wetter (2010-2012 Bainbridge: Z=6.1, N=1020, P<0.001; 2010-2012 Puerto Ayora:

Z=2.27, N=1037, P= 0.023), with significantly weaker winds (Z=-3.29, N=933,

P<0.001), than the 2012 sediment trap period (Figure D-7). As a result, the lake was significantly cooler (2012-2010: Z=-7.77, N=939, P<0.001) and more dense (2012-2010:

Z=3.63, N=939, P<0.001), during the 2012 sediment trap period than during the 2010 period (Figure D-8). These results suggest that the chemistry of the lake was driven primarily by the local temperature and precipitation budget, as there was no significant

184 difference in evaporation at the lake between the two sediment trap periods (2010-2012:

Z=1.8, N=923, P=0.071).

Water chemistry

Analysis of the chemical composition of water samples collected on 12/1/2009,

6/6/2010, and 10/3/2012 provide further support for wetter conditions in during the 2010 sediment trap collection period and drier conditions during the 2012 collection period

(Table D-2). The 2010 water sample was characterized by lower concentrations of cations and anions at the surface than at depth, particularly for Na, Si, Cl, and SO4. This surface depletion of major cations and anions at the surface is consistent with high input of freshwater to the lake through rainfall. The low surface silica is also consistent with high productivity at the lake during this period. Bicarbonate was also lower throughout the water column in 2010 than in either 2009 or 2012, providing further support for wetter conditions at the lake. In 2012, high bicarbonate concentrations and a lack of stratification of ions with depth are consistent with dry conditions and strong trade-wind driven vertical mixing of the lake. A charge imbalance between the major cations and anions in 2012 is consistent with a significant DIC concentration in the lake at this time.

Further, calcium and sulphate concentrations were lower than observed in either 2009 or

2010, consistent with the precipitation of carbonate and gypsum minerals in the lake.

Finally, silica was not depleted in the surface waters, suggesting that diatom productivity was low during this period.

185

Sedimentation

The composition and abundance of trace elements in the sediment trap sediments

(Figure D-11) were different between sampling periods and from that of the parent volcanic rock material (scoria). Microprobe mapping of a rock sample collected from the crater rim surrounding the lake demonstrated that the parent rock material was composed primarily of Si, Al, Fe, and Mg, along with some Na, Sr, and K (Figure D-9a). The μ-

XRF intensities of major and minor trace elements in five rock samples and one soil sample collected from the rim of the lake support the identification of the parent material as Fe, Si, Al and Mg rich (mafic) scoria (Swanson et al. 1974). In the soil sample, as well as in the rock samples with visible evidence of weathering relative to the parent rock material, the Al/Si was significantly higher than in the parent rock (Figure D-10;

H=216.3, N=495, P<0.001; Usoil-rock= 9.98, N=168; P<0.001), suggesting that Si is readily leached during weathering. The Bainbridge soil sample had properties of a torrand, an andisol common to aridic climates (Soil Survey Staff 2010) and observed in other parts of the Galápagos archipelago (Franz et al. 1980). The acidity of the lake water

(pH 6-7) likely results from weathering and subsequent rainwater flushing of the soil (soil pH 6.5).

The sediment trap samples from both years were rich in organic matter. Biogenic silica (diatom frustules) was a major component of the Si in both years, but was particularly prevalent throughout the 2012 sample. Biogenic silica comprised 4.9% of the

2010 sediment trap sample and 7.8-9.9% of the 2012 sediment trap sample. Microscopic investigation (20x to 30x) of 1 cc of sediment from each sediment trap supported higher content of organic matter in 2012 than in 2010, including both green algae and diatoms,

186 although this may be in part attributed to the longer sampling period which covered two warm seasons (when productivity in the lake is high).

In 2010, the sediment trap samples were dominated by Si, K, Na, and Mg (Figure

D-9b), with low levels of calcium. Sodium in these samples was found independent of chlorine and intensities of chlorine were low, suggesting that the salts had been sufficiently rinsed from the sediment during sample preparation. The occurrence of Na with Si, suggests that sodium may be associated with clay minerals in the sediments. In contrast to the low levels of calcium in the parent rock material and the 2010 sediments, calcium was abundant throughout the 2012 sediment trap samples (Figures D-9c-d). The median and distribution of μ-XRF Ca intensities was significantly different between the 2 sediment trap periods (2010 & 2012), with a higher median and more positively skewed distribution of Ca intensities in 2012 compared to 2010 (D=3.656, N=1630, P<0.001; Z=-

3.64, N=1630, P<0.001; Figure D-12). The high calcium intensities in the tail of this positively skewed distribution reflect the μ-XRF beam hitting carbonate crystals within the heterogeneous sediment sample that was otherwise low in calcium. Examining the crystallography of this carbonate material using BSE images and the EDS spectra on the microprobe revealed that the carbonate precipitating in the lake in both years was primarily composed of aragonite crystals, with low levels of Mg and Sr inclusion (Figure

D-13 g-i). In 2012, a <1 mm thick layer of carbonate also precipitated on the walls of the sediment traps (Figures D-9d & D-11). BSE imaging of these carbonates revealed a plate-like structure and aragonitic crystallography (Figure D-13 j-l).

The median and distribution of μ-XRF intensities were also different between years for several other elements. We hypothesize that higher intensities of Mg, Si, S, K,

187

Ti, V, Mn, Fe, and Sr in 2012 may be attributed to increased dust load to the lake driven by stronger winds during this period. Wind speeds measured at the lake were significantly greater during the 2012 sediment trap period than in the 2010 sediment trap period (Figure D-7); stronger winds during this period were also measured by the TAO buoy at 0ºN, 95ºW. The median wind direction recorded at Bainbridge Crater also shifted from E-ENE during the 2010 sediment trap period to S-SSE during the 2012 sediment trap period. Therefore, not only were the winds weaker during the 2010 sediment trap period, but the lake was also likely shielded from these winds by the high crater walls to the north during 2010.

Previous studies have demonstrated a significant input of atmospheric dust to the ocean in the eastern equatorial Pacific from the nearby dry continental margins (e.g.,

Prospero and Bonatti 1969; Scheidegger and Krissek 1982). The chemical composition of samples of atmospheric dust collected on a cruise southwest of the Galápagos archipelago (Figure D-1) by Prospero and Bonatti (1969) was distinct from the parent rock material at Bainbridge Crater (Figure D-14). Higher Na/K and Mg/Mn ratios were observed in the atmospheric dust sample and lake sediments than in the Bainbridge bedrock, suggesting a significant dust contribution to the elemental composition of the lake sediments (Figure D-14). The distribution of Na/K ratios was significantly different between sediment trap periods (D=1.57, N=1608, P=0.015), with a more positively skewed distribution in 2012 than in 2010. The distribution of Mg/Mn ratios was also more positively skewed in 2012 than in 2010, but the differences were not significant

(D=1.29, N=1544, P=0.07). These results provide support for a higher dust flux into the

188 lake in 2012 from the dry continental margins of South America as a result of strong southeast trade winds during this period.

Age model

Radiocarbon dating of the dissolved inorganic carbon in a water sample collected from 3 meters depth in 2012 suggests a reservoir age of 87 14C years ±35 (Table D-3), providing support for the contribution of both atmospheric and seawater sources to the lake carbon reservoir. Thus, we used a mixed calibration curve based on a reservoir age of 87 14C years (±35) to calibrate the radiocarbon dates from all of the cores. Based on this new information regarding the reservoir age of the lake, we also updated the chronology of the 1991 core accordingly (Riedinger et al. 2002) and tested the sensitivity of the chronology and laminae count to the inclusion of outlier radiocarbon dates at depths of 40.8, 264.8, 312.8, 321.7, and 333.55 cm (Figure D-15c, Figure D-S2). We present results from the age model with the outliers at 264.6, 312.8 and 321.7 cm depth removed (Figure D-15c), and use error bars to indicate the range observed across all age models. For the 2009 core, the southern hemisphere post-bomb curve from Hua and

Barbetti (2004) was used to calibrate the post-bomb radiocarbon date from 5.5-5.9 cm depth and the surface of the core was set to the 2009 collection date (Table D-3, Figure

D-15a). The radiocarbon date from 46.3 cm was an outlier (Table D-3) and was thus excluded from the age model. This sample fizzed profusely during acid pretreatment, suggesting that it had high carbonate content; thus, carbonate remaining after pretreatment may have led to the anomalously old age of this sample. The resulting age

189 models suggest that the 84 cm 2009 record was deposited in 815 years, the 290 cm 2007 record was deposited in 2294 years, and the 415.8 cm 1991 record was deposited in 6239 years (Figure D-15). There was no evidence of a hiatus in the sediment record from these cores. The average sedimentation rate was thus between 0.07 and 0.13 cm/year in

Bainbridge Crater Lake. However, considerable variability in sedimentation rate is observed through time (Figure D-15).

Climate history from Bainbridge sediment cores

All cores displayed two ~12 cm thick sections of coarse gypsum gravel at ~4-6 cm depth (1-8mm in size) deposited on top of alternating light (carbonate) and dark brown (siliciclastic) laminations, as observed by Riedinger et al. (2002) (Figure D-16).

Sediments between these distinctive siliciclastic and carbonate laminations was composed of a light brown to olive-green organic rich matrix. μ-XRF mapping of sediments from a strongly laminated section at ~71-76 cm illustrates these alternating

Si/Fe-rich and Ca/Sr-rich laminae (Figure D-17). The dominant mode of variability observed in singular value decomposition (SVD) of the μ-XRF elemental intensity correlation matrix (accounting for 32% of the variance), displayed strong positive loadings for S, Ca and Sr (eigenvectors: 0.52, 0.61, 0.41, respectively) and negative loadings for Si and Fe (eigenvectors: -0.32 and -0.25, respectively). Based on results from our ~3-year monitoring of the lake presented above, we interpret positive loadings in the first principal component (PC1) explaining the most variance in the XRF intensities to indicate dry conditions, when evaporites and carbonates formed within the

190 lake, and negative loadings in PC1 to indicate wetter conditions dominated by siliciclastic and terrigenous sedimentation.

Because positive PC1 values (high Ca, Sr and S intensities) bracket packets of gypsum gravel, we further propose that the packets of gypsum formed in the lake during prolonged dry periods when the lake became saturated with sulphate. High variability in grayscale values were also observed with these changes in sediment composition throughout the record (Figure D-18), with low to moderate values in the carbonate and gypsum sediments observed during dry intervals and high grayscale values in the brown- green organic-rich and siliciclastic sediments observed during wet intervals. Finally, we interpret high counts of nonlocal pollen grains to reflect a high influx of pollen from surrounding regions by the strong trade winds observed at Bainbridge during dry periods, whereas high counts of local pollen grains reflect the response of the vegetation on the crater (Sesuvium sp. and Croton sp.) to precipitation during wet periods.

Strong centennial to millennial-scale variability was observed in the composition of the Bainbridge sediment over the past six thousand years (Figure D-18-19). Low PC1 values and an abundance of local pollen grains suggest that wet conditions have persisted at Bainbridge since 0 cal. years BP (1950 CE). Positive PC1 values associated with gypsum layers (evaporites) and carbonate minerals and high abundance of regional pollen relative to local pollen suggests that La Niña-like conditions (dry with strong winds) persisted between ~350 and 0 cal. years BP (1600-1950 CE) and between 400-600 cal. years BP (1350-1550 CE). Around ~900 cal. years BP (1050 CE), we observe a large shift in the sedimentation regime at Bainbridge (Figure D-18). Prior to 900 cal. years BP

(1050 CE), the sedimentation rate in Bainbridge was much higher (1.5 mm/year

191 compared to 0.84 mm/year after 900 cal. years BP, Table D-3, Figure D-15) and lighter- colored, more variable sediments were deposited in the lake. This shift corresponds to an increase in the number of carbonate laminations within the lake, and a concurrent decrease in the number of siliciclastic laminations. This sediment regime, characterized by high and variable grayscale values and a peak in the number of carbonate laminations

(Riedinger et al. 2002), persisted from ~900 to 1500-1750 cal. years BP (200-1050 CE,

Figure D-18). Around ~1750 cal. years BP, another shift in sedimentation regime occurred within the lake, with moderate grayscale values and roughly equal numbers of carbonate and siliciclastic laminae observed between 2500 and ~1750 cal. years BP (550

BCE- 200 CE). Prior to 2500 cal. years BP, there were generally fewer laminations of both types within the Bainbridge sediment record, consistent with a weaker mid-

Holocene ENSO. Nonetheless, two peaks in the number of carbonate laminations occurred around 6100-5500 and 4000-3500 years BP (Table D-4, Figure D-19).

Discussion

In this work we present a new climate interpretation of the sediment record from

Bainbridge Crater Lake based on observations collected from long-term monitoring of the climate and limnology of the lake. Sediments collected from December 2009 to June

2010, a period consisting of a moderate El Niño event and warm, wet conditions, consisted primarily of brown-green, organic rich sediments, with very little authigenic carbonate. These sediments were therefore mainly composed of terrigenous elements found in the parent rock material (Si, K, Na, and Mg). Silica was abundant in the parent

192 rock material, but was rapidly leached from weathered rock material. In contrast, carbonate was abundant in the sediments collected between June 2010 and October 2012, a period that consisted of two full seasonal cycles and a La Niña event and that was significantly cooler and dryer. We therefore propose that the brown-green, organic rich, siliciclastic laminae in the Bainbridge sediment record reflect warm, wet conditions observed during El Niño events, when terrigenous material is washed into the lake from the crater walls and DIC concentration remains low. In contrast, we propose that carbonate laminae form during cool, dry conditions of neutral ENSO conditions and also

La Niña events, when DIC is concentrated in the lake water. Furthermore, we hypothesize that gypsum forms in the lake during prolonged dry periods when the lake water becomes saturated with sulphate. Although extended dry conditions were not observed over the ~3 year monitoring period presented here, high Ca, Sr and S intensities

(positive PC1 loadings) bracketing the intervals of gypsum gravel provides additional support for dry conditions during these intervals.

Applying this new climate interpretation to the Bainbridge Crater Lake sediment record, we find considerable variability in the frequency of both phases of ENSO throughout the last six thousand years. Comparing the Bainbridge sediment record with other regional reconstructions, we also discuss changes in overall state of the tropical

Pacific Ocean and ENSO-related variability within the basin. However, it is important to note the limitations of this comparison given age model uncertainty in these reconstructions. Age modeling the Bainbridge sediment record proved to be particularly challenging given the absence of macrofauna or flora to date and the presence of age reversals in the dates of bulk sediment samples. We therefore discuss climatic shifts

193 within the sediment record within this uncertainty, noting that the precise timing of such shifts may vary among records as a result of age model uncertainty.

The Bainbridge sediment record provides support for reduced ENSO variability during the mid-Holocene (Moy et al. 2002; Riedinger et al. 2002; Koutavas et al. 2006;

Conroy et al. 2008; Donders et al. 2008). With the exception of two periods around 6100-

5500 and 4000-3500 cal. years BP, low numbers of siliciclastic and carbonate laminations are observed in the Bainbridge sediment record between 6000 and 2500 cal. years BP. The moderate peak in siliciclastic and carbonate laminations around 6100-

5500 cal. years BP corresponds to a peak in the percent sand and silt at El Junco lake

(Conroy et al. 2008; Figure D-19), suggesting warm/wet conditions and relatively strong

ENSO-related variability in the Galápagos during this period. In contrast, the El Junco record suggests only a moderate number of El Niño events and generally cool/dry conditions between 4000-3500 cal. years BP, when a large peak in carbonate laminae number was observed at Bainbridge, suggesting that this peak in carbonate laminations resulted from more persistent La Niña-like conditions. The co-occurrence of gypsum within the Bainbridge sediment record during this interval is consistent with persistent dry conditions. This peak in carbonate laminations may have been associated with a strengthened SST gradient (Rein et al. 2005; Stott et al. 2004); however, the two eastern

Pacific SST reconstructions (Koutavas et al. 2002; Rein et al. 2005) disagree over this interval (Figure D-19). Nonethless, the presence of dry conditions at Bainbridge Crater lake during this interval is consistent with La Niña-like conditions that would result from a strong SST gradient (Rein et al. 2005).

194

Following generally low mid-Holocene levels of ENSO variability, the

Bainbridge sediment record suggests an intensification of ENSO variability starting around 2500 cal. years BP. Moderate grayscale values and roughly equal numbers of carbonate and siliciclastic laminae are observed at Bainbridge between 2500 and ~1750 cal. years BP (550 BCE- 200 CE), suggesting moderate ENSO-related variability during this interval. A marked increase in carbonate laminations and grayscale values within the

Bainbridge sediment record occurs around 1750 cal. years BP (200 CE). A peak in the number of carbonate laminations and high, variable grayscale values between ~1750 and

900 years BP suggest strong ENSO-related variability at Bainbridge crater lake during this interval. Other eastern Pacific paleoclimate reconstructions indicate similar climate shifts around 2000 years BP (Figure D-19), with an increase in the number of both El

Niño (Moy et al. 2002; Rein et al. 2005; Conroy et al. 2008; Makou et al. 2010; Seddon et al. 2011) and La Niña events (Makou et al. 2010), although the inferred timing of this shift varies among records. Based on the mean and uncertainty of the timing of this climate shift among the tropical Pacific climate reconstructions, a marked increase in

ENSO-related variability occurred around 1775 (± 190) cal. years BP. Alkenone and

Mg/Ca-based SST reconstructions from sediment cores in the eastern and western tropical Pacific suggest that these changes in ENSO variability were coeval with a reduction in the strength of the zonal SST gradient (Stott et al. 2004; Rein et al. 2005;

Conroy et al. 2008).

After ~900 years BP (~1050 CE), the sedimentation rate in Bainbridge lake slowed dramatically and the grayscale and laminae records indicate a shift towards generally wetter, less variable conditions at the lake (an increase in El Niño events, but

195 decrease in overall amplitude of ENSO-related variability). This transition occurred along with a strengthening of the zonal SST gradient (Stott et al. 2004; Rein et al. 2005) and suggests an overall decrease in the number of ENSO events relative the interval between 1750 and 900 cal. years BP. Deposition of gypsum and high regional pollen counts suggest that predominately dry, La Niña conditions persisted in the eastern equatorial Pacific between 400-600 BP (1350-1550 CE) and between ~350 BP and 0 BP

(1600-1950 CE) and relatively wetter, El Niño conditions persisted after 0 BP (1950 CE).

The packets of gypsum gravel within these extended dry intervals coincide with periods of solar minima observed during the Little Ice Age. A number of North American megadroughts have also been identified from tree-ring reconstructions during these intervals (Woodhouse and Overpeck 1998; Stahle et al. 2000; Stahle et al. 2007). These results suggest that the Little Ice Age (LIA) was generally drier in the Galapagos relative to the Medieval Climate Anomaly (MCA), consistent with a La Niña-like LIA as previously suggested by other hydrolocal reconstructions from the eastern equatorial

Pacific and Indo-Pacific warm pool (Oppo et al. 2009; Tierney et al. 2010; Yan et al.

2011). However, these hydrological records are in stark contrast to SST reconstructions, which suggest an El Niño-like LIA (Cobb et al. 2003; Newton et al. 2006; Conroy et al.

2009; Oppo et al. 2009). Although uncertainties in the interpretation of these paleoclimate proxy records may contribute to this discrepancy, the difference between

SST and hydrological proxy records suggests that the atmospheric and oceanic response to forcing in the tropical Pacific may be decoupled on centennial timescales. Further work is needed to investigate the mechanisms behind centennial-scale climate variability in the equatorial Pacific.

196

Conclusions

Based on results from long-term monitoring of Bainbridge Crater Lake, we propose that the brown-green, organic rich, siliciclastic laminae in the Bainbridge sediment record indicate warm, wet conditions typical of El Niño events, whereas carbonate laminae reflect cool, dry conditions typical of La Niña events. Based on this new climate interpretation, the Bainbridge sediment record suggests considerable variability in the frequency of both phases of ENSO throughout the last six thousand years. Consistent with other climate reconstructions from the tropical Pacific, the

Bainbridge sediment record suggests that ENSO variability was generally weak during the mid-Holocene, with moderate ENSO variability between 6100 and 5000 cal. years

BP. From 4000 to 3500 cal. years, abundant carbonate laminations were observed in the

Bainbridge sediment record, consistent with cool, dry conditions in the region that may have been associated with a strengthening of the tropical Pacific SST gradient. Starting around 2500 cal. years BP, the Bainbridge sediment record displays an increase in ENSO variability from mid-Holocene levels. The record suggests moderate ENSO-related variability between 2500 and ~1750 cal. years BP, followed by a large increase in the number of events around 1750 cal. years BP. ENSO-related variability in the Bainbridge record remained strong between ~1750 and ~900 cal. years BP, consistent with a weakened zonal SST gradient and strong ENSO variability in other regional records. The

Bainbridge sediment record therefore provides additional support for a marked increase in ENSO-related variability around 2ka. Based on estimates of the timing of this shift in the available tropical Pacific reconstructions, we estimate that ENSO variability

197 increased dramatically around 1775 (± 190) cal. years BP. The Bainbridge sediment record suggests that ENSO-related variability decreased again around ~900 cal. years BP as the zonal SST gradient strengthened. Over the past millennium, the Bainbridge sediment record suggests that persistent, La Niña-like conditions were observed between

400-600 BP and between ~350 BP and 0 BP (1950), while wetter, El Niño-like conditions were observed after 0 BP.

Acknowledgments

We thank S. Truebe, M. Miller, N. Dozouville, R. Pepolas, D. Ruiz, A. Tudhope,

M. Wilson, and C. Chilcot for providing assistance in the field and the Charles Darwin

Research Station and the Galápagos National Park for logistical support. We also thank

W. Kenney and M. Brenner for Pb-210 analyses. This research was supported by the

NSF RAPID and Atmospheric and Geospace Sciences Paleoclimate Programs (award

AGS-1256970), NOAA climate program, The University of Arizona Department of

Geosciences, and the Philanthropic Education Organization.

References:

Anhalzer JJ, Green JR (2011) Galapagos. Imprenta Mariscal, Quito, Ecuador, pp 95.

Battarbee RW, Kneen MJ (1982) The use of electronically counted microspheres in

absolute diatom analysis. Limnology and Oceanography 27:184–188.

198

Blauuw M (2010) Methods and code for ‘classical’ age-modelling of radiocarbon

sequences. Quaternary Geochronology 5:512-518.

Buol SW, Hole FD, McCracken RJ (1997) Soils genesis and classification. Second

edition. Iowa State University Press, Ames, Iowa, USA.

Clement AC, Seager R, Cane MA, Zebiak SE (1996) An Ocean Dynamical Thermostat.

Journal of Climate 9:2190-2196

Collins M, An SI, Cai WJ, Ganachaud A, Guilyardi E, Jin FF, Jochum M, Lengaigne M,

Power S, Timmermann A, Vecchi G, Wittenberg A (2010) The impact of global

warming on the tropical Pacific ocean and El Niño. Nature Geoscience 3: 391-397.

Conroy JL, Overpeck JT, Cole JE, Shanahan TM, Steinitz-Kannan M (2008) Holocene

changes in eastern tropical Pacific climate inferred from a Galápagos lake sediment

record. Quaternary Science Reviews 27(11-12): 1166-1180.

Conroy JL, Restrepo A, Overpeck JT, Steinitz‐Kannan M, Cole JE, Bush M, Colinvaux

PA (2009) Unprecedented recent warming of surface temperatures in the eastern

tropical Pacific Ocean. Nature Geoscience 2: 46‐50.

Charles Darwin Foundation (2012) CDF Meteorological Database - Base de datos

meterologico de la FCD. Online data portal - portal de datos en linea:

http://www.darwinfoundation.org/datazone/climate/ Last updated April 30, 2013

Deser C, Phillips AS, Alexander MA (2010) Twentieth century tropical sea surface

temperature trends revisited. Geophys. Res. Lett. 37: L10701,

doi:10.1029/2010GL043321.

199

Donders TH, Wagner-Cremer F, Visscher H (2008) Integration of proxy data and model

scenarios for the mid-Holocene onset of modern ENSO variability. Quaternary

Science Reviews 27(5): 571-579.

Faegri K, Iversen J (1989) Textbook of pollen analysis 4th Ed. Munksgaard, 237 p.

Franz H (1980) Old Soils and Land Surfaces on the Galápagos Islands. GeoJournal 4.2:

182-184.

Guilyardi E, et al. (2009) Understanding El Niño in ocean‐atmosphere general circulation

models: Progress and challenges. Bull. Am. Meteorol. Soc. 90(3): 325–340,

doi:10.1175/2008BAMS2387.1.

Guilyardi, E, H Bellenger, M Collins, S Ferrett, W Cai, and A Wittenberg (2012) A first

look at ENSO in CMIP5, Clivar Exchanges, 17(1): 29-32.

Hayes SP, Mangum LJ, Picaut J, Sumi A, Takeuchi K (1991) TOGA-TAO: A Moored

Array for Real-time Measurements in the Tropical Pacific Ocean. Bull. Amer.

Meteor. Soc. 72: 339–347.

Hua Q, Barbetti M (2004) Review of tropospheric bomb radiocarbon date for carbon

cycle modelling and age calibration purposes. Radiocarbon 46: 1273-1298.

Karnauskas KB, Seager R, Kaplan A, Kushnir Y, Cane MA (2009) Observed

Strengthening of the Zonal Sea Surface Temperature Gradient across the Equatorial

Pacific Ocean. J. Clim. 22: 4316-4321.

200

Kim, ST, and J-Y Yu (2012) The two types of ENSO in CMIP5 models, Geophys. Res.

Lett. 39: L11704, doi:10.1029/2012GL052006

Koutavas A, Lynch-Stieglitz J, Marchitto Jr. TM, Sachs JP (2002) El Niño-Like Pattern

in Ice Age Tropical Pacific Sea Surface Temperature. Science 297(5579): 226-230.

Koutavas A, deMenocal PB, Olive GC, Lynch‐Stieglitz J (2006) Mid‐Holocene El Nino‐

Southern Oscillation (ENSO) attenuation revealed by individual foraminifera in

eastern tropical Pacific sediments. Geology 34: 993‐996.

Liu Z, Vavrus S, He F, Wen N, Zhong Y (2005) Rethinking Tropical Ocean Response to

Global Warming: The Enhanced Equatorial Warming. J. Clim. 18:4684-4700.

Makou MC, Eglinton TI, Oppo DW, Hughen KA (2010) Postglacial changes in El Niño

and La Niña behavior. Geology 38(1): 43-46, doi: 10.1130/G30366.1

McPhaden MJ (1993) TOGA-TAO and the 1991-93 El Niño Southern Oscillation Event.

Oceanography 6(2): 36-44.

McPhaden MJ, et al, In: Proceedings of the “OceanObs’09:Sustained Ocean Observations

and Information for Society” Conference (Vol. 2), Venice, Italy, 21-25 September

2009, Hall, J., D.E. Harrison, and D.Stammer, Eds., ESA Publication WPP-306.

Meehl GA, et al. (2007), Global climate projections, in Climate Change 2007: the

physical science basis. Contribution of Working Group I to the Fourth Assessment

Report of the Intergovernmental Panel on Climate Change, edited by S. Solomon et

al., pp. 747–845, Cambridge Univ. Press, Cambridge, U. K.

Mitchell TP, Wallace JM (1992) The annual cycle in equatorial convection and sea

201

surface temperature. Journal of Climate 5: 1140-1156.

Moy CM, Seltzer GO, Seltzer DT, Anderson DM (2002) Variability of El Nino/Southern

Oscillation activity at millennial time scales during the Holocene epoch. Nature

420(6912): 162-165.

Newton A, Thunell R, Stott L (2006) Climate and hydrographic variability in the Indo‐

Pacific Warm Pool during the last millennium. Geophysical Research Letters 33(19).

Oppo DW, Y Rosenthal, BK Linsley (2009) 2,000-year-long temperature and hydrology

reconstructions from the Indo-Pacific warm pool. Nature 460(7259): 1113-1116

doi:10.1038/nature08233

Prospero JM, Bonatti E (1969) Continental dust in the atmosphere of the eastern

equatorial Pacific. Journal of Geophysical Research 74(13): 3362-3371.

Rasband WS (1999) ImageJ. US National Institutes of Health, Bethesda, MD, U.S.A.

http://rsb.info.nih.gov/ij/

Rein B, Lückge A, Lutz R, Sirocko F, Wolf A, Dullo C-W (2005) El Niño varia-bility off

Peru during the last 20,000 years. Paleoceanography 20: PA4003 (1-17),

doi:10.1029/2004PA001099

Riedinger MA, Steinitz-Kannan M, Last WM, Brenner M (2002) A similar to 6100 C-14

yr record of El Niño activity from the Galapagos Islands. J Paleolimnol 27(1): 1-7.

202

Scheidegger KF, Krissek LA (1982) Dispersal and deposition of eolian and fluvial

sediments off Peru and northern Chile. Geological Society of America Bulletin 93:

150-162.

Shuttleworth WJ (1993) Evaporation, in Maidment, D. R. ed., Handbook of Hydrology,

McGraw‐Hill, p. 4.1‐4.53.

Soil Survey Staff (2010) Keys to Soil Taxonomy, 11th ed. USDA-Natural Resources

Conservation Service, Washington, DC.

Stahle DW, Cook ER, Cleaveland MK, Therrell MD, Meko DM, Grissino-Mayer HD,

Watson E, Luckman BH (2000) Tree‐ring data document 16th century megadrought

over North America. EOS, Transactions American Geophysical Union 81(12): 121-

125.

Stahle DW, Fye FK, Cook ER, Griffin RD (2007) Tree-ring reconstructed megadroughts

over North America since AD 1300. Climatic Change 83(1-2): 133-149.

Stevenson, S, B Fox-Kemper, M Jochum, R Neale, C Deser, and G Meehl (2012) Will

there be a significant change to El Nino in the twenty-first century? Journal of

Climate 25: 2129-2145.

Stott LD, Cannariato KG, Thunell R, Haug GH, Koutavas A, Lund S (2004) Decline of

surface temperature and salinity in the western tropical Pacific Ocean in the Holocene

epoch. Nature 431:56-59.

203

Swanson FJ, Baitis HW, Lexa J, Dymond J (1974) Geology of Santiago, Rábida, and

Pinzón Islands, Galápagos. Geological Society of America Bulletin 85(11): 1803-

1810.

Tierney JE, Oppo DW, Rosenthal Y, Russell JM, Linsley BK (2010) Coordinated

hydrological regimes in the Indo‐Pacific region during the past two

millennia. Paleoceanography, 25(1).

Thompson DM, Ault TR, Evans MN, Cole JE, Emile-Geay J (2011) Comparison of

observed and simulated tropical climate trends using a forward model of coral δ18O.

Geophysical Research Letters 38, doi:10.1029/2011gl048224.

Trueman M, d’Ozouville N (2010) Characterizing the Galapagos terrestrial climate in the

face of global climate change. Galapagos Research 67: 26-36.

Vecchi GA, Clement A, Soden BJ (2008) Examining the tropical Pacific’s response to

global warming. Eos Trans. AGU 89(9), doi:10.1029/2008EO090002.

Wolff M (2010) Galapagos does not show recent warming but increased

seasonality. Galapagos Research 67: 38-44.

Woodhouse CA, Overpeck JT (1998) 2000 years of drought variability in the central

United States. Bulletin of the American Meteorological Society 79(12): 2693-2714.

Yan H, Sun L, Wang Y, Huang W, Qiu S, Yang C (2011) A record of the Southern

Oscillation Index for the past 2,000 years from precipitation proxies. Nature

Geoscience 4(9): 611-614.

204

Yeh, S-W, Y-G Ham, and J-Y Lee (2012) Changes in the tropical Pacific SST Trend

from CMIP3 to CMIP5 and its implication of ENSO. Journal of Climate 25: 7764-

7771.

205

Tables:

Table D-1: Bainbridge YSI profile collected from the deepest portion of the lake on 6/6/2010

Depth (m) Temp DO (%) Conductivity (mS)

0 29.4 90 139.1

1 29.3 83 138.9

2 29.1 78 138.7

3 28.8 75 138.3

206

Table D-2: Water chemistry of Bainbridge Crater Lake during 3 sampling periods: isotopic composition, pH, alkalinity, and major cations and anions. (*Finnigan Delta S in the Environmental Isotope Laboratory, run in February 2010. **Los Gatos Research, DLT- 100 (V1) instrument, run in October 2013. All other isotope values and replicates were performed on the Thermo Delta XP plus, Gas Bench II in May-June 2013. Replicates are indicated in parentheses).

Collection Date 12/1/09 6/6/2010 10/3/2012

Water Depth (m) 0 1 2 3 0 1 2 3 0 1 2 3

3.526 3.482 3.453 3.481 3.616 3.586 3.605 3.557 3.027 3.078 (3.582) 18O (‰) 3.093* 3.414* (3.665) (3.594) (3.715) (3.023) (3.213) (3.289) (3.381) 3.506* 3.286* 2.937** 3.19** 3.11** 3.09** 2.888** 3.388** 3.222** 3.35** 3.85** 3.602**

13 15 15 15 D (‰) 10.836** 10.742** 10.545** 9.477** 10.537** 9.959** 11.208** 12.184** 12.664** 11.257**

pH 6-6.5 6-6.5 6-6.5 6-6.5 6 6 6 6 7 7

- HCO3 (mM) 2.35 2.34 2.49 2.28 1.80 1.78 1.81 1.78 2.30 2.25

Ca (mg l-1) 1333.92 1230.24 1274.72 1184.54 1016.73 1071.50 1128.49 1118.29 992.16 971.16

Mg (mg l-1) 3982.27 3670.04 3864.31 3504.01 3010.94 3199.95 3333.35 3440.65 2984.73 2856.66

Na (mg l-1) 32856.59 31483.09 32654.78 30955.62 26558.05 27853.56 28658.82 29201.96 25415.79 25431.6

K (mg l-1) 1209.83 1175.66 1215.84 1181.04 1018.34 1066.96 1073.84 1099.12 982.85 978.35

Sr (mg l-1) 23.17 21.57 22.47 20.86 17.69 18.72 19.47 19.95 17.29 17.0

207

Si (mg l-1) 15.24 14.00 14.71 14.70 12.44 13.56 13.40 14.21 13.25 13.14

B (mg l-1) 19.47 17.64 17.91 17.18 14.70 15.56 15.75 16.31 14.92 14.39

Cl (mg l-1) 55785.3 56216.5 57899 53310.2 47400.2 49310.7 50651.0 51762.8 44544.7 43307.6

SO4 (mg l-1) 6844.8 6925.0 7234.0 6392.3 5684.3 6035.7 6272.7 6463.8 5485.6 5278.4

Br (mg l-1) 190.8 191.5 196.0 183.2 162.8 170.8 175.6 175.4 154.7 148.0

Sum of cations (charge 1854.37 1762.89 1833.08 1724.14 1479.75 1555.63 1604.65 1637.24 1425.77 1414.76 x mM)

Sum of anions (charge x 1720.74 1734.57 1788.67 1641.34 1459.17 1520.46 1563.28 1598.59 1374.89 1335.54 mM)

Balance 100x[cations- 3.74 0.81 1.23 2.46 0.7 1.14 1.31 1.19 1.82 2.88 anions]/[cations+anions]

208

Table D-3: 14C measurements and ages of Bainbridge material.

13 core Sample Depth Lab  C 14C age FMC /sample Description (cm) number (‰) (years)

Bain2B Bulk sed 5.5-5.9 AA101362 -14.0 1.1645±0.0049 Post-bomb

Bain2B Bulk sed 17.8-18.4 AA101363 -8.9 0.9194±0.0041 675±35

Bain2B Bulk sed 31.3-32.5 AA101364 -10.5 0.9089±0.0040 767±35

Bain2B Bulk sed 46.3-47.0 AA101365 -12.9 0.8541±0.0039 1,266±36

Bain2B Bulk sed 53.3-54.8 AA101366 -12.9 0.8736±0.0039 1,086±36

Bain2B Bulk sed 57.7-58.7 AA101367 -10.9 0.8971±0.0040 872±36

Bain2B Bulk sed 66.4-67.8 AA101368 -11.8 0.8804±0.0040 1,023±36

Bain2B Bulk sed 71.2-72.3 AA101369 -11.7 0.8807±0.0039 1,021±35

Bain2B Bulk sed 76.8-77.5 AA101370 -12.0 0.8762±0.0043 1,062±39

Bain2B Bulk sed 82.0-83.0 AA101371 -11.7 0.8765±0.0039 1,059±35

3m 2012 DIC N/A AA101514 -11.3 0.9892±0.0043 87±35

209

Table D-4: Number of carbonate and siliciclastic laminae in the 1991 core from Bainbridge Crater Lake. The Reidinger et al. (2002) record was reanalyzed in 250 year bins from 0 to 2500 years BP and in 500 year bins from 2500 to 6000 years BP based on the new age model for the core developed in this work.

Years BP Siliciclastic Carbonate 0-250 25 0 250-500 10 2 500-750 4 28 750-1000 4 42 1000-1250 2 21 1250-1500 0 47 1500-1750 5 16 1750-2000 8 6 2000-2250 16 7 2250-2500 5 5 2500-3000 1 12 3000-3500 7 13 3500-4000 0 67 4000-4500 2 3 4500-5000 5 3 5000-5500 2 1 5500-6000 14 21 6000-6100 0 31

210

Figures:

(a) (b)

(c)

(b)

Figure D-1: (a) Map of study region showing location of Bainbridge Crater Lake and the sites from which data were obtained for this study: meteorological stations at Bainbridge (this study) and Puerto Ayora (Charles Darwin Foundation, 2012), Tropical Atmosphere Ocean (TAO) array moored buoy at 0ºN95ºW, and the approximate cruise track on which a dust sample was collected by Prospero and Bonatti (1969). (b) Aerial photo (Anhalzer and Green 2011) of Bainbridge Crater Lake showing the location of the weather station (A) and the deepest portion of the lake (~3.3 meters, B) where sondes and sediment traps were deployed and where sediment cores were collected in 2009 and 2012. (c) Bathymetric map of the lake with 0.5 meter contour lines.

211

Figure D-2: Time series of local climate and limnology for the study period: 12/2/2009 to 10/3/2012. (a) sea-surface temperatures (SSTs, ºC) from Puerto Ayora (black dashed, Charles Darwin Foundation 2012) and a TAO array station at 0ºN, 95ºW (Hayes et al. 1991; McPhaden 1993; McPhaden et al. 2009; gray), (b) air temperature recorded by the Bainbridge weather station (black) and water temperature at 1m (blue solid) and 2m (blue dashed) depth recorded by AquaTroll 100 sondes, (c) average daytime solar radiation measured from the weather station, (d) wind speed recorded by the weather station (black) and by the TAO array at 0ºN, 95ºW (gray), (e) wind direction (degrees) recorded by the weather station (black) and by the TAO array at 0ºN, 95ºW (gray), and (f) precipitation (mm/day) at Bainbridge (black) and Puerto Ayora (black dashed) and salinity (PSU) of the lake at 1m (blue solid) and 2m (blue dashed) depth. The two sediment trap periods (12/2/2009-6/6/2010 & 6/6/2010-10/2/2012) are indicated at the bottom of the figure, and the warm seasons (January 1 to May 31) are highlighted in red.

212

Figure D-3: Histogram of daily average rain (left), air temperature (left center), wind speed (right center) and evaporation (right) for the warm seasons (January 1- May 31, top) and cool seasons (June 1- December 31, bottom) between December 2009 and October 2012. The median value is denoted by a white dotted line on each histogram.

213

Figure D-4: Histogram of daily average salinity (left), water temperature (left center), density (right center), and total dissolved solids (TDS, right) for the warm seasons (January 1- May 31, top) and cool seasons (June 1- December 31, bottom) between December 2009 and October 2012. The median value is denoted by a white dotted line on each histogram.

214

Figure D-5: Histograms of daily average rain (left), air temperature (left center), wind speed (right center) and evaporation (right) for the 2010 (top), 2011 (middle) and 2012 (bottom) warm seasons (1/1-5/31). The median value is denoted by a white dotted line on each histogram.

215

Figure D-6: Histogram of daily average salinity (left), water temperature (left center), density (right center), and total dissolved solids (TDS, right) for the 2010 (top), 2011 (middle) and 2012 (bottom) warm seasons (1/1-5/31). The median value is denoted by a white dotted line on each histogram.

216

Figure D-7: Histogram of daily average rain (left), air temperature (left center), wind speed (right center) and evaporation (right) for the 2010 sediment trap sample (12/2/2009-6/5/2010, top), and 2012 sediment trap sample (6/6/2010-10/2/2012, bottom). The median value is denoted by a white dotted line on each histogram.

217

Figure D-8: Histogram of daily average salinity (left), water temperature (left center), density (right center), and total dissolved solids (TDS, right) for the 2010 sediment trap sample (12/2/2009-6/5/2010, top), and 2012 sediment trap sample (6/6/2010-10/2/2012, bottom). The median value is denoted by a white dotted line on each histogram.

218

(a) (b)

Figure D-9: BSE image and maps of Ca, Sr, Si, Fe, Ti, K, S, Mg, P, Mn, and Na in a sample of the (a) parent rock material from the rim of Bainbridge Crater Lake, and sediments collected at 2m depth between (b) December 2, 2009 and June 6, 2010 and (c- d) June 6, 2010 and October 2, 2012. Sample (c) was collected from a ~1cm carbonate rich lamination between the top of the test tube and the overlying bottle funnel, estimated to contain ~80% brown, organic rich seds, and ~20% carbonate. Sample (d) was collected from a <1mm thick carbonate layer covering the interior of the sediment trap bottle wall. The parent rock material was composed mainly of Si, Al, Fe, and Mg, while the sediments were composed of primarily of Si, K, Na, and Mg in 2010 and Ca, Sr, Si, Na, Mg in 2012.

219

Figure D-10: Boxplot of the Al/Si ratio in 3 rock samples, 2 partially weathered rock samples, and 1 soil sample taken from the rim of the crater lake.

220

Figure D-11: Images of the sediment traps deployed at Bainbridge Crater Lake at 1m depth (B, E) and 2m depth (A,C,D,F) upon recovery on June 6, 2010 (top) and on October 2, 2012 (bottom). Images A & D show the sediments that were collected in and along the sides of the funnel above the centrifuge collection tubes (C, F).

221

Figure D-12: Histogram of calcium intensity measured on the μ-XRF in sediment trap samples collected from 2m depth in (a) 2010 (sediments accumulating between 12/2/2009 and 6/6/2010) and (b) 2012 (sediments accumulating between 6/6/2010 and 10/2/2012).

222

Figure D-13: BSE images and WDS spectra of (a-c) the parent rock material, scoria; (d- f) diatom frustules; (g-i) aragonite crystals; and (j-l) plate-like carbonate and aragonite crystals that lined the walls of the 2012 sediment trap.

223

Figure D-14: Log-log scatter plot of the ratio of Mg/Mn and Na/K intensities in the parent rock material (red), 2010 sediment trap samples (light blue) and 2012 sediment trap samples (blue) measured on the μXRF. The ratio of Mg/Mn and Na/K measured in atmospheric dust collected in a cruise southwest of the Galapagos Islands (Prospero and Bonatti 1969) is shown for comparison in black.

224

(a) (b)

(c)

Figure D-15: Bainbridge clam (Blauuw 2010) age model for (a) the 360 cm core collected in 2007, (b) the 92 cm core collected in 2009 and (c) the 415.8 cm core collected in 1991 (Riedinger et al. 2002). Gray area highlights maximum and minimum age estimates for core (95% confidence intervals for 100,000 iterations) and black line indicates best-fit (median) age model. Outliers that were excluded from the age model are highlighted in red, and the surface dates of the cores (set to the date of coring) are highlighted in green. Additional dates tagged as outliers to test for the sensitivity of the 1991 core chronology to outliers are circled in black. A total of 7 age-depth models with varying number of outliers excluded were developed for this core (see Supplemental Figure 2), and the one shown here was selected for further analyses.

225

Figure D-16. The Bainbridge sediment record and radiocarbon ages from cores collected in 2007 (left), 2009 (middle) and 1991 (right, Riedinger et al. 2002).

226

Figure D-17: Image and XRF maps of elemental intensities of Core Bain2B from ~71-76 cm depth plotted in colorized grayscale. Lighter colors indicate higher abundance of the elements.

227

Figure D-18: (a) PC1 of µ-XRF ratios of Si/K, S/K, Cl/K, Ca/K, Ti/K, Fe/K and Sr/K in Bainbridge core 2B collected in 2009 (32% of the variance), (b) Bainbridge sediment grayscale record measured on the 2007 long core, where high grayscale values indicates lighter colors, (c) Bainbridge pollen counts in the 2007 core: total pollen count, local pollen (Sesuvium sp. and Croton sp.) and regional (other) pollen count, (d) Carbonate and siliciclastic laminae (Riedinger et al. 2002) counted in 250 year bins, (e) Percent sand grains in the El Junco sediment record (Conroy et al. 2008), (f) Red color intensity as a 18 proxy for terrestrial runoff in Laguna Pallcacocha, Ecuador (Moy et al. 2002), (g) δ Osw reconstruction from δ18O and Mg/Ca of planktonic foraminifera in marine sediment cores from the Indo-Pacific warm pool (IPWP) (Oppo et al. 2009), and (h) δD value of leaf waxes from marine sediment cores from the West Sulawesi margin, IPWP (Tierney et al. 2010). Histograms in (d) are centered on the midpoint of the bin, and histogram error bars and confidence intervals (for total laminae) denote the range observed across all 7 age models tested. The major periods of solar minima are outlined in gray and the approximate timing of the Medieval Climate Anomaly (MCA) and Little Ice Age (LIA) are denoted.

228

Figure D-19: (a) Western tropical Pacific sea-surface temperature (SST) derived from foraminiferal Mg/Ca in a core from 5°00.18'S, 133°26.69'E in 2382 m water depth (red;

229

Stott et al. 2004) and composite cores from the Makassar Strait, Indonesia (light red; Oppo et al. 2009) (b) Eastern tropical Pacific SST derived from alkenones in a core from the Peru margin (blue, 12°030’S, 77°39.80’W in 184 m water depth, Rein et al. 2005) and from foraminiferal Mg/Ca in a core from 1°13’S, 89°41’W in 617 m water depth (light blue, Koutavas et al. 2002) (c) PC1 of µ-XRF ratios of Si/K, S/K, Cl/K, Ca/K, Ti/K, Fe/K and Sr/K in Bainbridge core 2B collected in 2009 (32% of the variance), (d) Bainbridge sediment grayscale record measured on the 2007 long core, where high grayscale values indicates lighter colors, (e) Bainbridge pollen counts in the 2007 core: total pollen count, local pollen (Sesuvium sp. and Croton sp.) and regional (other) pollen count, (f) Carbonate and siliciclastic laminae (Riedinger et al. 2002) counted in 500 year bins, where error bars and dotted lines are as defined in Figure 20, (g) Number of El Niño events s in 100-yr overlapping windows measured from a sediment core from Laguna Pallcacocha (Moy et al. 2002), (h) Lithic flux (% of max) in a core from the Peru margin (12°030’S, 77°39.80’W in 184 m water depth, Rein et al. 2005), (i) Percent silt grains in the El Junco sediment record (Conroy et al. 2008), (j) as in (i) for percent sand grains (Conroy et al. 2008), (k) Abundance of dinosterol, a sterol produced by dinoﬂagellates, in the organic carbon (OC) contained in a sediment core from 11°4'S, 78°5'W in 252 m depth (Makou et al. 2010), (l) as in (k) for the abundance of cholesterol, a eukaryotic sterol that reflects total primary productivity and zooplankton activity in the surface waters (Makou et al. 2010), and (m) El Junco Lake, Galápagos tychoplanktonic to epiphytic (T:E) diatom SST index (Conroy et al. 2009).

230

Figure D-S1: Time series of local limnology at 1 meter depth recorded by AquaTroll 100 sondes (black) over the full study period: 12/2/2009 to 10/3/2012 and water samples (blue dots) collected on 12/1/2009, 6/6/2010, and 10/3/2012. (top) water temperature at 1m (blue solid), (middle) specific conductivity (mS/cm) and (bottom) salinity (PSU).

231

232

Figure D-S2: Sensitivity test of the clam (Blauuw 2010) age model for the 415.8 cm core collected in 1991 (Riedinger et al. 2002) to radiocarbon date outliers. Gray area highlights maximum and minimum age estimates for core (95% confidence intervals for 1,000 iterations) and black line indicates best-fit (median) age model. Outliers that were excluded from the age model are highlighted in red, and the surface date (set to the date of coring) is highlighted in green. A total of 7 age-depth models with varying number of outliers excluded were developed for this core (including the one selected for further analysis, Figure 15), and the results of this sensitivity test were used to calculate uncertainty estimates for the laminae counts from this core.

233

APPENDIX E: PERMISSIONS

234

APPENDIX A: COMPARISON OF OBSERVED AND SIMULATED TROPICAL CLIMATE TRENDS USING A FORWARD MODEL OF CORAL 18O, is reprinted with permission from John Wiley and Sons

Thompson, D.M., T.R. Ault, M.N. Evans, J.E. Cole, and J. Emile-Geay, (2011). Comparison of observed and simulated tropical climate trends using a forward model of coral δ18O. Geophys. Res. Lett., 38, L14706, doi:10.1029/2011GL048224.