Reconstructing a Centennial-Scale Extreme Paleoflood History of the Pee Dee River Using Oxbow Lake Sediments

Coastal Carolina University CCU Digital Commons

Electronic Theses and Dissertations College of Graduate Studies and Research

8-7-2019

Reconstructing a Centennial-scale Extreme Paleoflood History of the Pee Dee River using Oxbow Lake Sediments

Nicholas William Conway Coastal Carolina University

Follow this and additional works at: https://digitalcommons.coastal.edu/etd

Part of the Geology Commons, and the Geomorphology Commons

Recommended Citation Conway, Nicholas William, "Reconstructing a Centennial-scale Extreme Paleoflood History of the Pee Dee River using Oxbow Lake Sediments" (2019). Electronic Theses and Dissertations. 113. https://digitalcommons.coastal.edu/etd/113

This Thesis is brought to you for free and open access by the College of Graduate Studies and Research at CCU Digital Commons. It has been accepted for inclusion in Electronic Theses and Dissertations by an authorized administrator of CCU Digital Commons. For more information, please contact [email protected].

Reconstructing a Centennial-scale Extreme Paleoflood History of the Pee Dee

River using Oxbow Lake Sediments

Nicholas W. Conway

Submitted in Partial Fulfillment of the

Requirements for the Degree of Master of Science in

Coastal and Marine Wetland Studies in the

School of the Coastal Environment

Coastal Carolina University

2019

Dr. Zhixiong Shen Dr. Robert Sheehan Major Professor Committee Member

Dr. Eric Wright Committee Member

Dr. Richard Viso Dr. Michael H. Roberts SCE Director Dean

Dedication This thesis is dedicated to my Mom, Dad, and Uncle Jeff. Without your love and support this work would have been impossible.

iii

Acknowledgements

I would like to express utmost gratitude to my major advisor Dr. Zhixiong Shen, and my committee members Dr. Robert Sheehan and Dr. Eric Wright. Their support of my work has helped me succeed as a graduate student and this thesis wouldn’t be here if it weren’t for them. A special thanks goes to Dr. Zhixiong Shen, who’s determination, scientific rigor, and seemingly endless intuition have pushed me to dig deeper and strive for the highest quality of science possible. Not only have you helped me become a better scientist, you have helped me become a professional and a more well-rounded individual. I also would like to thank Dr. Erin Hackett for providing insight in data analysis and helping to extensively improve my scientific computing skills.

Thank you to Dr. James Tompkins and the radiology technicians at Conway Medical

Center for providing CT scans for this research. Another thanks goes to Dr. Samuel Muñoz at the Marine Science Center at Northeastern University, Dr. Barbara Mauz at the

University of Salzburg, and the National Ocean Sciences Accelerator Mass Spectrometer

Facility at Woods Hole Oceanographic Institute for providing dating measurements.

Thank you to my funding sources that made this research possible: Coastal Carolina

University and the Douglas D. Nelson Fellowship. Additionally, I would like to thank

Richard Goldberg and Sam Gary for allowing us to use CCU’s boats during field work.

And finally, I would like to extend a huge thanks to my family, my friends, and my girlfriend. Your support is never ending, and I am utterly grateful to have you all in my life.

Abstract

Extreme river floods are the key force shaping floodplain landscape and a major process delivering sediment, pollutants, and nutrients to coasts. These devastating natural hazards pose concerns about potential change of extreme flood occurrence in the face of climate change. However, accurately assessing the impact of anthropogenic climate change and natural climate modes on the intensity and frequency of extreme flooding relies on multi- century discharge records. Unfortunately, instrumental records are relatively short (often

<100 years) and overlap with times of dam and reservoir construction. Oxbow lakes, ubiquitous in the floodplains of alluvial rivers, may preserve an archive of extreme flood at centennial timescales as they capture coarser channel sediments transported by intensified river flows. This study has identified signals of extreme floods in oxbow lake sediments and established a timeline of past flooding events to evaluate change(s) in flood hazard near the Pee Dee River (PDR), South Carolina.

Laser diffraction grain-size analysis and X-ray computed tomography (CT) scanning were performed on a ~2-m long piston core (SBL2) to identify event layers of extreme floods.

CT images reveal high-density laminations and corresponding coarser shifts of grain size are interpreted as flood layers. A robust age-depth model was established for SBL2 using multiple independent age controls (C14, optically stimulated luminescence, Pb210/C137, and historical event tie-points). End-member modelling analysis was performed to identify a coarse component of the grain-size data used as a proxy of extreme flood. A linear relationship between end-member modelling results and measured discharge was established for the last 80 years and applied to the older part of the core yielding peak discharge estimates back to ca. AD 1840.

This analysis identifies abrupt shifts in grain size resulting from dam construction, droughts, and local geomorphic changes to the river system. A multidecadal trend in the frequency of extreme floods is present in the PDR system, controlled by Pacific Decadal

Oscillation. The most extreme peak annual discharges of the PDR occurred between AD

1870-1900 from the combined interaction of increased tropical cyclone activity with intensified land use for agricultural purposes. Peak annual discharges of the PDR seem to have decreased through time since flood control damming was completed in AD 1962.

List of Tables

Table 1. List of macrofossils used for radiocarbon dating. Table 2. Dating samples used in this study and their corresponding depths.

vii

List of Figures

Figure 1. Southeastern United States showing rivers and USGS gauge stations. USGS gauge Pee Dee River at Pee Dee, SC is marked near the 6 on the map. Adapted from Leigh et al. (2004). Figure 2. Flood history of the PDR at Pee Dee, SC (USGS Gauge 02131000) for 1938- 2018. A. Peak annual gauge height measurements and river stage categories are displayed. B. Peak annual discharge measurements are displayed. The black dotted line represents the discharge limit of the paleoflood detection method designed in this study (see Methods section). Peaks above the National Weather Service’s major flood stage in A. are labelled with their corresponding date of occurrence in B.. Figure 3. Topographic map of the PDR with oxbow lakes labeled. Inset map shows major Atlantic coast rivers flowing through Virginia, North Carolina, South Carolina, and Georgia. The small gray box in the inset map highlights my study area. Figure 4. Stages of abandonment for a meander bend cutoff which forms an oxbow lake. Adapted from Toonen et al. (2012). Figure 5. An infrared satellite image from the National Aeronautics and Space Administration showing the meteorological conditions that contributed to the 2015 record rainfall event in SC. A frontal boundary can be seen crossing the entire length of SC in dark red and black, a large upper level low pressure system is seen as light blue over Georgia and Florida, and Hurricane Joaquin is situated over the Atlantic Ocean to the south funneling moisture towards SC. Figure 6. Observed AMO index AD 1870-2015, defined as detrended 10-year low pass filtered annual mean area-averaged SST anomalies over the North Atlantic basin (0N-70N, 80W-0E), using HadiSST dataset (Rayner et al. 2003) for the period AD 1870-2015. Obtained from Trenberth & Zhang (2019). Figure 7. Spatial extent of the PDO (top) and ENSO (bottom) SST variability and their respective time-series for the period AD 1870-2014. Positive phase of the PDO (top-left) and ENSO (bottom-left) are displayed. Obtained from Deser & Trenberth (2016). Figure 8. Power spectra of the normalized PDO and ENSO time-series. While periods at the 5-year scale contribute some variance to the PDO time-series, periods of ~20- 50 years contribute the dominant amount of variance. For the ENSO time series periods at the 5-year scale contribute the most variance. Figure 9: Map of dam locations (blue pins) along the Pee Dee-Yadkin river system in NC, and the study area of South Balloon Lake (yellow pin). Image courtesy of Google Earth. Figure 10. Location of South Balloon Lake sediment core (SBL2) used in this study. South Balloon Lake is an oxbow lake formed by neck cutoff of the Pee Dee River, SC. Bathymetric contours (black) are given in feet.

viii

Figure 11: SBL2 core photograph. Figure 12: The age-depth model for SBL2 shows the median age probability (black line) and 1σ confidence intervals (bracketed by red-dotted lines), with 1σ error bars on individual chronological controls. Figure 13: End member modelling outcomes identified three dominant end members in SBL2. Figure 14: Grain-size distribution of an extreme paleoflood deposit (115 cm). Figure 15. Grain-size distribution of background sediments (87 cm) deposited during non- flooding conditions. Figure 16. CT-scan image, EM1 score, z-scored EM1, and CT number for SBL2 reveal coarse mineral-rich layers used as tracers of extreme paleofloods. A. The coarsest EM (EM1 score) has change points (gray dotted lines) at 20 cm and 66 cm that may correspond to a dirt road construction in the floodplain and the beginning of dam construction, respectively. B. Z-scores that exceed the extreme flood threshold (gray dotted line) have been assigned to flood years that fall within the age-depth models 1σ uncertainty range. C. CT data confirm flood events identified in z-scored EM1 data, however some events are overcounted. Figure 17. Scatterplot and linear regression with 1σ prediction intervals relating flood layer z-score to peak annual discharge measurements of the Pee Dee River at the USGS gage station at Pee Dee, SC. Figure 18. Peak annual discharge estimates (AD 1842 – 1945) with 1σ prediction intervals of paleoflood layers. Paleoflood discharges were estimated using the linear relationship between flood-bed z-score and instrumental peak annual discharge measurements. Peak annual discharges from AD 1945 to AD 2018 are instrumental peak annual discharges. The red bar at the top shows a period of more intense floods, and the blue bar shows a period of decreasing flood magnitude. Figure 19. PDO index and PDR flood frequency (31-year window) exhibit a positive correlation that is statistically significant (Pearson’s R=0.39, p<0.001). Figure 20. AMO index and PDR flood frequency exhibit weakly positive correlation that is not statistically significant (Pearson’s R=0.14, p>0.05). Figure 21. Correlation between major winter/spring flood events (black stars) of the PDR with Multivariate ENSO Index Version 2 calculated using JRA-55 reanalysis (Kobayashi et al., 2015). Figure adapted from NOAA ESRL (2019, https://www.esrl.noaa.gov/psd/enso/mei/). Figure 22. Correlation between the most extreme floods caused by hurricanes (black stars) and AMO index. Adapted from Trenberth and Zhang (2019). Figure 23. Composite anomaly of 500mb geopotential height over North America during El Niño events. Trough in geopotential height is evident across the southern tier of the US. Image courtesy of NOAA’s Earth System Research Laboratory.

Figure 24. Effects on general temperature, wetness, and storminess of North America during El Niño winters. Figure 25. Composite precipitation anomalies over the US during El Niño winters. Positive precipitation anomalies exist over the PDR catchment. Image courtesy of NOAA’s Earth System Research Laboratory. Figure 26: Chance of extreme precipitation over the US during El Niño winters (Left: November, December, January; Right: December, January, February). Extreme precipitation is defined as being in the wettest 20% of the 100-year record. Wet extremes are +90% increase in risk for the majority of the PDR catchment during El Niño events. Image courtesy of NOAA’s Earth System Research Laboratory. Figure 27: Composite anomaly of precipitation rate (mm/day) over the US during El Niño winters. Increased precipitation rate is observed over the PDR catchment area in NC and SC. Image courtesy of NOAA’s Earth System Research Laboratory. Figure 28: Google earth maps showing a dirt road constructed between AD 1986 and AD 1987 near SBL2. The white arrow on the right photo points to the road. Figure 29. Number of tropical storms and hurricanes affecting the PDR basin from AD 1851-2018. Figure 30. Storm track of a hurricane in AD 1878. Data obtained from NOAA Office of Coastal Management. Figure 31. Storm track of a hurricane in AD 1896. Data obtained from NOAA Office of Coastal Management. Figure 32. Storm Track of a hurricane in AD 1928. Data obtained from NOAA Office of Coastal Management. Figure 33. Storm track of a Hurricane in AD 1945 that caused the largest peak annual discharge in the instrumental record on September 22nd, 1945. Data obtained from NOAA Office of Coastal Management. Figure 34. Level of erosive land use in the southern Piedmont from AD 1700-1967. Figure adapted from Trimble (2008). Figure 35. Peak annual discharge of the Saluda River at Chappels, SC. Figure 36. Peak annaual discharge of the Congaree River at Columbia, SC. Figure 37. Peak annual discharge of the Savannah River at Burtons Ferry bridge near Mill Haven, GA. Figure 38. Peak annual discharge of the Edisto river near Givhans, SC. Figure 39. Peak annual discharge of the Waccamaw River near Longs, SC. Figure 40. Precipitation map of the historic SC rainfall event of 2015. Adapted from CISA (2016).

List of Abbreviations US: United States ENSO: El Niño Southern Oscillation PDO: Pacific Decadal Oscillation AMO: Atlantic Mutlidecadal Oscillation SEACP: Southeast Atlantic Coastal Plain PDR: Pee Dee River NC: North Carolina SC: South Carolina CT: Computed Tomography USGS: United States Geological Survey AD: anno Domini SST: Sea Surface Temperature EM: End Member RoC: Rate of Change SBL: South Balloon Lake OSL: Optically Stimulated Luminescence CT90: 90% Cumulative Frequency of CT Number

TABLE OF CONTENTS

Dedication…...…..…………………………………………………………….…………iii

Acknowledgements……………………………………………………………………...iv

Abstract…………………………………………………………………………………...v

List of Tables……………………………………………………………………………vii

List of Figures…………………………………………………………………….…….viii

List of Abbreviations………………………………………………………………….…xi

1. Introduction……………………………………………………………………………1

2. Study Area……………………………………………………………………….……..5

3. Literature Review………………………………………………………………….…11

3.1 A synopsis of modern floods in the southeastern US and their causes….....11

3.2 Previous paleoflood studies……………………………………………….…14

3.3 Changes of major climate modes during the late Holocene………………....17

3.4 Land use and river management change of the PDR system………………..22

3.5 Paleoflood research methods………………………………………………..24

4. Methods……………………………………………………………………………….28

4.1 Field Methods………………………………………………………………..28

xii

4.1.1 Bathymetry Survey…………………………………………………28

4.1.1 Sediment Coring…………………………………………………...29

4.2 X-ray CT scanning…………………………………………………………...29

4.3 Lab methods………………………………………………………………….30

4.3.2 Grain-size analysis…………………………………………………30

4.3.3 Radiocarbon dating………………………………………………..31

4.3.4 210Pb and 137Cs dating……………………………………………...31

4.3.5 Optically Stimulated Luminescence……………………………….32

4.4 Flood Identification and Paleodischarge Estimate…………………….…...32

4.5 Age-depth Modelling……………………………………...…………………33

4.6 Flood Frequency Analysis……………..…………………………………….34

5. Results………………………………………………………………………………...35

5.1 Core Sedimentology………………………………………………………….35

5.2 Core Chronology………………………………………………...…………..36

5.3 Grain-size and CT data…………………………………………………..….38

5.4 Paleodischarge Estimates……...………………………………………...…..41

5.5 Climate Modes and Flood Frequency……………………………………….43

6. Discussion……………………………………………………………………………..46

xiii

6.1 Paleoflood Detection Methods……………………………………….……...46

6.2 Climatic and Anthropogenic Controls on Flood Frequency………………..46

6.2.1 Summer & Fall Floods………………..…………………………..47

6.2.2 Winter & Spring Floods……………………………………………49

6.3 Change of Peak Annual Discharge………………………..………….……..54

7. Conclusions…………………………………………………………………………...63

Tables……………………………………………………………………………………64

References……………………………………………………………………………….66

xiv

1. Introduction

Prediction that anthropogenic warming may increase Northern Hemisphere total precipitation by 30% creates concerns about future flood frequency and intensity

(Dankers at al., 2014; Paasche & Støren, 2014). Total and extreme precipitation increase, however, represent two distinct aspects of precipitation with the latter being more likely to cause extreme floods. Observations and climate models show robust increases of global extreme (annual maximum) daily precipitation over the past six decades and climate projections suggest continued intensification of extremes during the rest of the century

(Donat et al., 2016; Wuebbles et al., 2017). This intensification has implications for the risk of flooding in many major United States (US) rivers.

Although no study has connected changes in flooding of major US rivers to anthropogenic climate forcing (Wuebbles et al., 2017), there is a large body of evidence showing that decadal climate modes, such as El Niño Southern Oscillation (ENSO), Pacific

Decadal Oscillation (PDO), and Atlantic Multidecadal Oscillation (AMO), and atmospheric circulation patterns affect the occurrence of flood favorable storms. Decadal climate modes are shown to have a profound effect on the timing and magnitude of extreme paleofloods globally (Patton & Dibble, 1982; Ely, 1992,1997; Wolfe et al., 2006; Wilhelm et al., 2012, 2013; Glur et al., 2013; Paasche & Støren, 2014; Muñoz et al., 2018). An understanding of climate modes and their effect on late-Holocene riverine flooding in the southeastern US, however, is poor, with only a few studies addressing this topic (Enfield

1 et al., 2001; Maleski & Marinez, 2018). Several studies trace millennial-scale paleofloods in the southeastern US through the Holocene (Wang & Leigh, 2012; Goman & Leigh,

2017), however, they are coarse in resolution and magnitudes of paleofloods are not resolved.

Regional setting plays an important role in flood occurrence, but greater frequency of intense floods in one region does not necessarily equate to greater frequency of intense floods everywhere. For example, studies performed in the European Alps, only ~100-200 km away from each other, yield opposite results when correlating frequent flooding to past warm/cool periods (Glur et al., 2013; Wilhelm et al., 2012, 2013). Similarly, regional differences exist in the average climatic conditions associated with extreme floods in the southwest US. The Arizona River tends to favor cool-wet climate periods during most extreme flooding (Ely, 1992; 1997), while the Pecos River in southwest Texas tends to favor arid climate conditions (Patton & Dibble, 1982).

A recent study by Muñoz et al. (2018) showed that artificial channelization and extreme precipitation related to climate modes, greatly amplify flood magnitudes of the

Lower Mississippi River. River engineering and channel modifications are common practices undertaken on many US rivers to store potable water supplies, control floods, provide recreation, and create hydroelectric power (Morton, 2003). Climate modes that impact flood favorable conditions of the Mississippi Basin (ENSO and AMO) affect weather patterns over the southeast Atlantic coastal plain (SEACP) as well, but they may not control extreme flood frequency and magnitude in both regions the same. To determine the driving force of extreme flooding of rivers in the SEACP more research is needed.

This study uses the Pee Dee River (PDR) basin to address questions about the driving force of extreme riverine flooding in SEACP during the late Holocene. This drainage basin carries water through agricultural regions of the Carolinas and to the Grand

Strand, an economically developing coastal vacation destination and residential area. Flood damage threatens these pivotally important economic entities, both of which generate necessary revenue for state and local governments of North Carolina (NC) and South

Carolina (SC). Additionally, large floods of the PDR put homes and lives at risk and disrupt logging, mining, and farming industries along the river. During floods, rivers within the

PDR basin can transport pollutants adsorbed to fine sediments, posing a threat to natural resources located in estuaries along the Atlantic Ocean. Building flood mitigation infrastructure and repairing flood damages is extremely costly. These costs lead many SC natives and policy makers to question when a deadly flood of unforeseen magnitude will occur, and if modern society should implement new adaptive measures. Unearthing the history of flooding along the PDR may reveal information needed to optimize flood mitigation strategies in this rapidly developing region.

The research questions addressed in this study include: Is natural climate variability the key force driving extreme flooding in the SEACP? Have dams and reservoirs constructed along rivers in the SEACP through the 20th century decreased the potential for extreme flooding downstream? Is anthropogenic climate change leading to more frequent, or more intense floods?

Answering these questions, unfortunately, is hampered by a lack of long-term hydrological data. A relatively short instrumental period (1938 – Present) overlaps with times of land-use change and dam construction, which have significantly modified the

PDR’s natural hydraulic and hydrologic behavior; and make it difficult to examine natural variability prior to these changes. This study advances our understanding of flood controls in the SEACP by generating a long-term paleoflood history that extends beyond current historical records.

An innovative method of computed tomography (CT) scanning and grain-size analysis of oxbow-lake sediments in the PDR floodplain is employed to quantify extreme paleofloods. The primary goal of this research is to produce a continuous event-scale paleoflood record, which will be useful for investigating changes in extreme flood frequency and intensity in the SEACP. Based on land-use change, river management, and natural climate variability, changes in the intensity and frequency of PDR floods are expected. Hypotheses for this research include: (1) The PDO causes a multidecadal trend in extreme flood frequency of the PDR. (2) During the period of most erosive land use,

1860-1920, more intense floods occurred. (3) Dam construction completed in 1962 has decreased peak annual discharge of major floods.

2. Study Area

The PDR basin is used for this study because it is among the largest river basins in the southeastern United States (Figure 1). The drainage area of the PDR at the United

States Geological Survey (USGS) gauge at Pee Dee, SC (02131000) is 22,870 km2. The

PDR begins in NC as the Yadkin River and becomes the PDR just south of the confluence of the Uwharrie and Yadkin rivers in Badin, NC. This river system flows from the base of the Blue Ridge Mountains, through the Piedmont, and across the Coastal Plain of SC emptying into Winyah Bay near Georgetown, SC. The present pattern of this river is characterized by meandering channels. The PDR transports bed loads of sand and fine to medium gravel and abundant suspended loads as most of its Piedmont sourced drainage is on clayey saprolite (Leigh et al., 2004).

The primary and most consistent flood season of the PDR is during the winter

(December-April), when heavy precipitation occurs mostly in connection with frontal passages associated with extratropical cyclones. The second flood season of the PDR occurs in late summer and fall. The summer/fall flood season is less consistent than the winter one, but there is a potential for high magnitude floods as they are generally caused by extreme precipitation of landfalling hurricanes.

Instrumental discharge measurements used in this study were obtained from the

USGS gauge, Pee Dee River at Pee Dee, SC (02131000). Flood categories for this site are defined by the National Weather Service Advanced Hydrological Prediction Service. Flood

5 stage is 5.8 m, moderate flood stage is 7.0 m, and major flood stage is 8.5 m. When classifying major flood events at this site the peak discharge threshold is ~2000 m3/s; the

National Weather Service, however, officially classifies floods by river stage not by discharge because channel characteristics are constantly changing, so an 8.5 m water level at a different time likely had a significantly different discharge then one measured for a modern flood (Richard Neuherz-NOAA, personal communication, July 2, 2019). Flood stage elevation, however, is measured as height above a fixed datum, which for this site, is

7.17 m above NAVD88. Over the instrumental record period, five major flood events have occurred at this site, with two caused by hurricanes. The two floods due to hurricanes, however, have the most extreme discharges measured.

The specific study site for this project is along the PDR floodplain near Gresham,

SC, where many oxbow lakes are present (79 ̊ 15’ – 24’ W, 33 ̊ 45’ – 55’ N) (Figure 3).

Oxbow lakes act as sediment traps during extreme flood events and can be used for paleoflood research. Oxbow lakes form by meander cutoff (Figure 4; Toonen et al., 2012) and fill by plug bar formation from the lake arms towards the center of the abandoned channel, overbank sedimentation, and authigenic locally derived sediments. Flood sediments are transported into oxbow lakes through tie-channels in plug bars during the onset of increased river discharge when an extreme flood begins.

Figure 1. Southeastern United States showing rivers and USGS gauge stations. USGS gauge Pee Dee River at Pee Dee, SC is marked near the 6 on the map. Adapted from Leigh et al. (2004).

B. 9/22/1945

9/21/2018

03/02/1979 03/07/1987 04/15/2003

Figure 2. Flood history of the PDR at Pee Dee, SC (USGS Gauge 02131000) for 1938- 2018. A. Peak annual gauge height measurements and river stage categories are displayed. B. Peak annual discharge measurements are displayed. The black dotted line represents the discharge limit of the paleoflood detection method designed in this study (see Methods section). Peaks above the National Weather Service’s major flood stage in A. are labelled with their corresponding date of occurrence in B..

Figure 3. Topographic map of the PDR with oxbow lakes labeled. Inset map shows major Atlantic coast rivers flowing through Virginia, North Carolina, South Carolina, and Georgia. The small gray box in the inset map highlights my study area.

Figure 4. Stages of abandonment for a meander bend cutoff which forms an oxbow lake. Adapted from Toonen et al. (2012).

3. Literature Review

3.1. A synopsis of modern floods in the southeastern US and their causes

In coastal areas of the southeastern US, riverine floods are often caused by heavy precipitation associated with tropical cyclones, extratropical cyclones, frontal passages, and intense convective thunderstorms (Hirschboeck, 1991). While in mountainous regions, orographic effects and snow melt commonly cause floods. Other factors influencing flooding include antecedent soil moisture and changes in land cover in urban or farmed areas. Along the SEACP, flooding occurs due to its relatively flat landscape that drains multiple major rivers (Cook, 1936).

Factors such as frozen ground and rapid snowmelt may exacerbate flooding caused by winter extratropical cyclones tracking northeastward along the east coast. During spring, large volumes of precipitable water vapor and high density contrasts between converging air masses can cause heavy rainfall. In summer, convectional processes and large amounts of precipitable water vapor sourced from the Atlantic Ocean and Gulf of Mexico, produce thunderstorms with intense precipitation of short duration, sometimes causing flash floods in smaller drainage basins (Hirschboeck, 1991). Heavy fall precipitation occurs primarily in connection with tropical cyclones, early winter storms, or stalled boundaries (Mizzell and Simmons, 2015).

In early October 2015, a combination of meteorological factors caused a record rainfall event in SC (Figure 5). A cold front pushed across the state from the west and

11 stalled off the US southeast coast as a strong high-pressure ridge formed over eastern

Canada. Simultaneously, Hurricane Joaquin strengthened to a category 4 hurricane over the Bahamas and a slow-moving upper low was held in place over Georgia by the high- pressure ridge to the north. A steady flow of tropical moisture associated with Hurricane

Joaquin was funneled into SC causing record precipitation and extreme flooding in multiple drainage basins. Maximum precipitation in the state was 68.3 cm in Mt. Pleasant,

SC (Musser et al., 2016).

More recently, SC has experienced coastal and inland flooding caused by hurricanes (e.g., Matthew in 2016; and Florence in 2018). Precipitation from Hurricane

Florence caused the highest gage height of 9.7 m at USGS Gage 02131000 on the PDR since 1945. Slow moving Hurricane Florence produced persistent rainfall that caused extensive flooding, lasting weeks after the storm. So far, these floods (2015, 2016, and

2018) have cost an estimated $37.2 billion and caused 127 deaths (NCEI, 2018).

Large river basins, such as the PDR, drain great volumes of floodwater and can take weeks to subside. The Hurricane Florence flood halted operations at Coastal Carolina

University in Conway, SC for three weeks, and damaged thousands of homes throughout

Horry County, SC. In order to prevent major highway 501 from being overtopped by record breaking floodwaters of the Waccamaw River, a wall was constructed out of concrete and sandbags (personal observation, 2018).

NC SC GA

Figure 5. An infrared satellite image from the National Aeronautics and Space Administration showing the meteorological conditions that contributed to the 2015 record rainfall event in SC. A frontal boundary can be seen crossing the entire length of SC in dark red and black, a large upper level low pressure system is seen as light blue over Georgia and Florida, and Hurricane Joaquin is situated over the Atlantic Ocean to the south funneling moisture towards SC.

3.2. Previous paleoflood studies

Paleoflood hydrology formally began with the work of Kochel and Baker (1982) along the lower Pecos and Devils rivers in southwestern Texas as reviewed by Baker

(2008). Baker (1987) explained that the most accurate technique for tracing paleofloods is the analysis of slackwater deposits and paleostage indicators, resulting from high velocity floods that characterize narrow, deep canyons in resistant geological materials. Baker

(2008) elaborated on the geographic and geologic controls of slack-water deposit analysis, indicating that rivers appropriate for this method commonly include bedrock canyons that transport sandy sediments. Due to these controls most studies have been conducted in the southwestern US (Kochel and Baker, 1982; Ely, 1985, 1997; Baker, 1987, 2008) with limited paleoflood hydrology studies in the southeast US (Wang & Leigh, 2012; Goman &

Leigh, 2017; Leigh, 2018). Rivers in this region have had major changes to their dimensions in the Holocene (Leigh et al., 2004; Leigh, 2008), countering traditional recommendations for slack-water deposit analysis (Koechel et al., 1982). The SEACP possess mostly sandy and silty unconsolidated sediments that are subject to frequent erosion. A different method, therefore, must be developed to trace paleofloods in the

SEACP.

Recent studies (Wolfe et al., 2006; Støren et al., 2010; Dietze et al., 2013) show that lake sediments often preserve characteristic flood layers and are commonly used for developing continuous sequences of past extreme flood events. Most of these studies, however, focus on high-altitude proglacial lakes or high-latitude temperate lakes that capture floods of relatively small distributary rivers (Støren et al., 2010; Glur et al., 2013;

Wilhelm et al., 2013; Schillereff et al., 2016). These studies are important to determine

European climate dynamics. Mountainous geographic regions are typically sparsely populated, however, and flood damage does not pose an immediate danger to communities.

Rarely studied are low-relief alluvial rivers, whose floodplains and deltaic plains are often heavily populated (Toonen et al., 2015; Muñoz et al., 2015, 2018). Flood damage in these regions threatens lives and causes costly damage. Alluvial rivers can change their path during floods leaving abandoned channels disconnected from the main channel, forming oxbow lakes. During normal river stage oxbow lakes continuously accumulate fine-grained organic-rich background sediment. During floods, coarse clastic-rich sediment load transported by increased flow from the main channel may be brought to oxbow lakes. Flood sediments are larger in grain size, less organic rich, and have greater bulk density in comparison with background sediment, leaving event layers that may be detected in oxbow lake stratigraphy.

A study conducted by Wolfe et al. (2006) used magnetic susceptibility and grain- size analysis of oxbow lake sediment cores to create a multi-century flood record of the

Peace River, Canada. This study showed that flood frequency had been in decline since the early 19th century preceding regulation of the Peace River in AD 1968. The study also presented evidence that several prolonged multi-decadal intervals (20 + years) have occurred without a flood including AD 1975-1995, AD 1840-1875, AD 1813-1839, and

AD 1705-1786. These findings provide evidence that frequency of Peace River floods is controlled primarily by natural climate variability rather than anthropogenic modification to the river system.

Glur et al. (2013) showed that frequent floods in the Swiss Alps coincide with cooler periods throughout the past 2500 years. They attribute this correlation to the

15 occurrence of westerly storm tracks and the Vb cyclone track (subcategory in a group of storm track pathways classified by roman numerals; Van Bebber, 1891) during cooler summers. The Vb cyclone track takes a northeastward path from the Adriatic Sea to the

European Alps (Messmer et al., 2015), bringing orographic induced rainfall with a potential for severe flooding. If a warming climate continues to yield warmer summers, European flood occurrence is expected to be lower than today (Glur et al., 2013). Glur et al. (2013) concluded that their findings are regional by nature, so paleoflood studies in different regions may yield different results. They also emphasized that it is important to consider the effects of spatial changes in atmospheric circulation when reconstructing flood records and projecting global warming climate scenarios.

In the Southwestern U.S. paleoflood studies have also shown that high magnitude floods are related to periodic changes in atmospheric circulation caused by climate modes

(Ely, 1997). Ely (1997) explained that times of increased frequency of high-magnitude floods coincide with cool-wet climate in the southwest U.S., that is conducive to highly meridional flow in the upper atmosphere and a southerly shifted storm track. Furthermore, frequent or persistent El Niño conditions were shown to cause more winter storms and dissipating tropical cyclones, which produce the most extreme floods in the modern flood record of rivers in the southwest U.S. (Ely, 1997).

Muñoz et al. (2018) linked multidecadal changes in flood hazard of the Mississippi

River to ENSO and AMO, with the magnitude of these floods being greatly amplified by anthropogenic river modifications. High magnitude floods of the Mississippi River tend to occur during negative AMO phases that produce extreme precipitation events over most of the Mississippi River basin, and discharge extremes are exceedingly large when they fall

16 on moist soils left by El Niño events (Muñoz et al., 2018). Muñoz et al. (2018) supports older studies performed on the Mississippi River, showing that the frequency and magnitude of Holocene floods are sensitive to only modest changes in climate (Knox, 1993,

2000). This research shows that high resolution flood data can be collected from oxbow lake sediments in rivers that are at high risk for flood damage.

In the southeast US a paleoflood investigation has been performed using sediments in the Upper Little Tennessee River valley (Wang & Leigh, 2012). Wang and Leigh (2012) found two periods in the last 2000 years with large floods during AD 650-850 and AD

1100-1350, which are times when regional tree-ring records showed extreme wetness.

Additionally, in the coastal plain of NC a pollen record from a peat core in a paleochannel of the Little River (Goman & Leigh, 2004) provided a pollen record and a record of overbank sedimentation for the past 10,500 years. Goman & Leigh (2004) concluded that increases in moisture and flood events between 9000 to 6100 cal yr BP were probably controlled by shifts in the position of the Bermuda High, sea surface temperatures, and El

Niño activity. Neither of these studies, however, have resolved discharge magnitudes of paleofloods events.

3.3. Changes of major climate modes during the late Holocene

It is important to understand that both external forcing and internal variability contribute to changes in global temperature and weather. External forcing causes climate perturbations and originates from energy sources beyond the natural climate system, including greenhouse gas inputs (global warming), solar variabilities, and aerosol injections. Internal variability describes transfers of energy that occur naturally in earth’s climate system. Heat transfers between ocean, atmosphere, and land are recorded by

17 indexes such as ENSO (Philander, 1983; Steinman et al., 2015), AMO (Delworth & Mann,

2000; Kerr, 2000; Enfield et al., 2001; Gray, 2004; Mann et al., 2014; Steinman et al.,

2015), and Pacific Decadal Oscillation (PDO) (Mantua et al., 1997; Zhang et al., 1997;

Desser et al., 2010, 2012; Desser & Trenberth, 2016).

Evidence of multidecadal sea surface temperature variation in the North Atlantic and its effects on Sahel rainfall, began with the work of Folland et al. (1984, 1986).

Subsequent modelling efforts by Delworth et al. (1993, 1997) demonstrated the existence of a multidecadal “stadium wave” pattern of variation, associated with thermohaline circulation and coupled ocean-atmosphere processes in the North Atlantic. The AMO is indexed as the 10-year low pass filtered annual mean area-averaged sea surface temperature (SST) anomalies in the North Atlantic (0-65°N, 80°W-0°E). AMO is constructed using the HadiSST dataset (Rayner et al., 2003) for the period AD 1870-2015

(Trenberth & Zhang, 2019) (Figure 6), and exhibits a ~50 to 70-year oscillation (Kerr,

2000; Gray et al., 2004). Using tree-ring proxies, AMO records have been extended as far back as AD 1567 (Gray et al., 2004). Instrumental records show that the AMO had warm phases at AD 1860-1900, AD 1930-1960, and AD 1998-present, and cool phases from AD

1905-1925 and AD 1970-1990 (Enfield et al., 2001; Gray et al., 2004; Trenberth & Shea,

2006; Trenberth & Zhang, 2019). Frajka-Williams et al. (2017) showed that a marginally negative AMO phase may have emerged since AD 2015 and is expected to persist and possibly intensify.

One major reason that floods and anthropogenic climate change have not been linked is the limited understanding of natural climate variability and its direct effect on floods. A study conducted by Steinman et al. (2015) showed that variability of natural

18 climate modes may suppress or amplify anthropogenic temperature increase in the northern hemisphere. This finding suggests that northern hemisphere weather patterns, such as heavy precipitation events, may also be suppressed or amplified by natural climate mode variability. Previous research demonstrated the effects of AMO on continental US rainfall

(Enfield et al., 2001; Muñoz et al., 2018). They showed that rainfall is positively correlated with AMO phase in central and southern Florida (Enfield et al., 2001), and negatively correlated in the Lower Mississippi River catchment (Enfield et al., 2001; Munoz et al.,

2018). Between AMO warm and cold phases, Mississippi River outflow varies by 10%, while inflow to Lake Okeechobee varies by 40% (Enfield et al., 2001). Mann and Emanuel

(2006) showed that there is a strong relationship between tropical Atlantic SST and tropical cyclone activity through the late nineteenth century. They stated that this relationship is not connected to the AMO, however, but more likely to the influence of anthropogenic forced warming. Goldenberg et al. (2001) argue that reduction of vertical wind shear associated with warmer SST and lower sea level pressure are in phase with positive AMO indices. Warm AMO phase is likely responsible for observed changes in major hurricane landfalls on the US east coast, which were low from AD 1903-1925 and AD 1971-1994, and high from AD 1926-1970 and AD 1995-2000 (Goldenberg et al., 2001; Zhang &

Delworth, 2006).

ENSO and PDO are indexes that explain SST anomalies in the tropical Pacific, and the north Pacific, respectively (Figure 7). Positive ENSO phases known as El Niño events are associated with positive SST anomalies in the eastern tropical Pacific. Negative phases are known as La Niña events, resulting in positive SST anomalies in the western tropical

Pacific. Positive PDO phases are associated with positive SST anomalies in the eastern

North Pacific and negative SST anomalies in the central and western north Pacific. The positive phase of the PDO is also associated with positive SST anomalies in the central and eastern tropical Pacific, and for this reason this reason PDO and ENSO have similar impacts on atmospheric circulation and weather over North America. The main difference between these two modes are their timescales. ENSO is an interannual to interdecadal phenomenon, while the PDO operates predominantly on decadal scales (Figure 8) (Deser

& Trenberth, 2016). Research shows that the positive phase of PDO is associated with an enhanced frequency of El Niño events (Verdon & Franks, 2006). This means that the PDO can be used as a multidecadal surrogate of El Niño event frequency.

Figure 6. Observed AMO index for the period AD 1870-2015, defined as detrended 10- year low pass filtered annual mean area-averaged SST anomalies over the North Atlantic basin (0N-70N, 80W-0E), using HadiSST dataset (Rayner et al. 2003). Obtained from Trenberth & Zhang (2019).

Figure 7. Spatial extent of the PDO (top) and ENSO (bottom) SST variability and their respective time-series for the period AD 1870-2014. Positive phase of the PDO (top-left) and ENSO (bottom-left) are displayed. Obtained from Deser & Trenberth (2016).

Figure 8. Power spectra of the normalized PDO and ENSO time-series. While periods at the 5-year scale contribute some variance to the PDO time-series, periods of ~20-50 years contribute the dominant amount of its variance. For the ENSO time series, periods at the 5-year scale contribute the most variance.

3.4. Land use and river management change of the PDR system In SC, the PDR is a free-flowing river, but the river has been dammed for hydroelectric power in 4 locations and for flood control in one location in NC (Figure 9).

The hydroelectric power dams were constructed between AD 1905 and AD 1928, while the northernmost dam built for flood control was constructed in AD 1962. These reservoirs, from northwest to southeast, are: W. Kerr Scott Dam and Reservoir, High Rock Lake,

Badin Lake, Lake Tillery, and Blewett Falls Lake. Dams can have a large effect on reducing downstream sediment load and water flow (Phillips, 1991; McCarney-Castle et al., 2010).

Through time land use has changed in the PDR catchment, particularly in the

Piedmont region. Trimble (2008) has extensively studied soil erosion on the southern

Piedmont for the period AD 1700-1967, stating that during the few centuries prior to

European settlement soil erosion was minimal. Clear cutting forests and farming for cotton and tobacco began to cause greater levels of erosion beginning in the mid to late AD 1700s

(Trimble, 1974). Intensity of erosive land use increased through the AD 1800s, with the most intense period occurring between AD 1860-1920. After which, this intensity greatly decreased between AD 1920-1967 with the decline of agriculture, the increase of conservatory land use, and an improvement of soil conservation methods for farming

(Trimble, 1974). Damming and erosive land use may have had impactful yet opposite effects on the amount of sediment the PDR transports to Winyah Bay. Some sign of this change should be detectable in oxbow lake sediments of the PDR in the SC coastal plain.

Figure 9: Map of dam locations (blue pins) along the Pee Dee-Yadkin river system in NC, and the study area of South Balloon Lake (yellow pin). Image courtesy of Google Earth.

3.5. Paleoflood research methods

Lithological properties of sediments are commonly used in sedimentary geology to interpret past or present environments. Lithological properties include sediment color, grain-size, sorting, and composition, and can be interpreted to determine sediment source, distance travelled, environmental conditions (e.g. anoxic/oxic), and the amount of energy needed for transport. These interpretations can lead to an understanding of past environments. Since major floods produce tremendously energetic river flow, they leave a unique lithologic footprint in oxbow lake sediments that can be used to create a record of past hydrologic extremes.

Studies have employed end member (EM) modelling of grain-size distributions as a statistical method to determine unique sediment transport processes of lake sediments

(Dietze et al., 2014; Yu et al., 2016). EM modelling programs gather all grain-size measurement results from a sediment archive, and unmixes subpopulations by isolating commonalities identified throughout the dataset using mathematical concepts of principal component analysis and factor analysis ([Dietze et al., 2012; Dietze & Dietze, 2016; Yu et al., 2016). These grouped commonalities, or EM, can be interpreted as proxies of sediment transport processes. EM modelling aids in the interpretation of flood events when analyzing a vertical sediment core, with coarser EM representing flood sediment transport pathways, and finer EM representing background sedimentation. EM modelling also calculates EM score, which is the percentage that a certain EM contributes to a given grain-size distribution.

Toonen et al. (2015) used EM modelling to analyze lake sediments and determine magnitudes of extreme past flood events on the lower Rhine River. They computed z-

24 scores (a measure of how many standard deviations below or above the population mean a raw score is) of coarse-tail EM score to normalize the data between trend breaks that exceed a 95% confidence level, identified using change-point analysis (Taylor, 2000). A linear regression between peak annual discharge and z-scores of corresponding flood beds in the instrumental period was used to estimate paleoflood discharges. This regression yielded a high correlation, R2 = 0.8, leaving 20% of the variance unexplained. The author justifies this unexplained variance with factors that affect how floods are registered among their research locations including local geomorphic dynamics, formation of river ice, river management, uncertainties in age-depth modelling, and uncertainties in the correlation between grain-size data and discharges.

Munoz et al. (2018) adapted the EM method by Toonen et al (2015) but excluded the z-scores. Instead, they normalized the coarsest EM using a low-pass (41 cm) moving minimum filter to remove long-term trends in sediment composition caused by local geomorphic processes. They then identified potential flood deposits as the normalized EM scores that exceeded a high-pass (11 cm) moving average by a 0.1 EM score threshold and compared identified flood deposits with their X-ray fluorescence and radiography data for verification. Linear models that describe normalized EM scores as a function of measured discharge were built to reconstruct peak annual discharge along the Lower Mississippi

River. Much more variance was left unexplained when reconstructing flood magnitudes in this study as compared to Toonen et al. (2015), with R2 values of 0.301, 0.345, and 0.603 for three different core locations by Muñoz et al. (2018).

Another approach in paleoflood studies is using computed tomography (CT) scanning of lake-sediment cores. Originally designed for medical use (Houndsfield, 1972),

CT scanners are rarely applied to paleoflood studies. However, applications to marine sediment cores (Orsi et al., 1994) and oil-reservoir rock cores (Withjack et al., 1990) determined lateral variation in sediment bulk density, and helped analyze oil flow processes, respectively. CT scanning is a non-destructive analysis tool that can assist previously established paleoflood detection methods by indicating variability of sediment bulk densities with depth (Orsi et al., 1993; Duliu, 1999; Støren et al., 2010).

CT scanners use X-rays as their source of energy, and X-ray detectors to detect the energy transmitted. When an object is placed between the source and the detectors, X-rays are attenuated. By manipulating the geometry between the source, object, and detectors, usually by continuous rotation of the X-ray tube as the object is mechanically moved through the beam (applies to 3rd generation CT scanners, e.g., see Storen et al., 2010), CT machines can provide a map of the X-ray attenuation at sub-millimeter slice intervals. For example, the Toshiba Aquilion 64 slice multi-slice spiral CT scanner sub-divides each slice into a matrix of 512 × 512 voxels (volume elements), and pixel values (CT numbers) are assigned to each voxel, providing a three-dimensional view of the core’s CT numbers expressed in Houndsfield units (HU) (Storen et al., 2010). The linear X-ray attenuation coefficient is directly proportional to bulk density (Orsi et al., 1994). The machines outputted HU scores are linearly related to the X-ray attenuation coefficient by:

HU(x,y) = [(µ(x,y) - µw) / µw] * N

where HU(x,y) is the computed Houndsfield unit as a function of position, µ(x,y) is the X-ray attenuation coefficient of the material, also as a function of position, µw is the linear attenuation coefficient of water, and N is a scaling constant equal to 1000 (adapted from,

Orsi et al., 1994). When viewing CT images less dense regions are shown as darker areas, and denser regions are lighter.

Storen et al. (2010) has successfully applied CT analysis to high-resolution lake sediment cores retrieved from southern Norway and produced a detailed record of

Holocene flood activity for the past 10,000 years. The record identified 317 events by using an objective approach, which calculates the rate of change (RoC) in CT-number and identifies events as those exceeding a one standard deviation threshold. RoC is calculated by dividing the change in parameter (dy) by change in time (dt):

RoC = dy/dt

A positive increase in the RoC is expected during the onset of a flood event, as CT number rapidly increases in response to increased flood-transported minerogenic sediments to the lake. As flood intensity and deposition rate gradually decrease the RoC becomes negative.

4. Methods

4.1. Field Methods

4.1.1. Bathymetry Survey

A bathymetry survey of South Balloon Lake was conducted using a Garmin echomap chirp 74dv to identify the thalweg of the paleochannel where the largest amount of lacustrine sediment should have accumulated (Figure 10). In addition to coring in the thalweg, we chose a site close to the initial river cutoff site to ensure a high contrast between background and flood sediments.

Figure 10. Location of South Balloon Lake sediment core (SBL2) used in this study. South Balloon Lake is an oxbow lake formed by neck cutoff of the Pee Dee River, SC. Bathymetric contours (black) are given in feet.

4.1.2. Sediment Coring

South Balloon Lake was selected for coring because it has remained stationary since at least AD 1939, according to historical aerial photographs. Following the geophysical survey, a coring location was selected near the plug-bar (cf. Toonen et al.,

2012), or the area nearest the initial channel cutoff site (at the time of oxbow lake formation), to increase the likelihood of high contrast between background and flood sedimentation (Munoz et al., 2018). A 2.05 m-long piston core (SBL2) was collected on

December 15, 2017 from a small aluminum boat. The core was wrapped in an opaque black trash bag to prevent light exposure in preparation for optically stimulated luminescence

(OSL) dating, returned to Coastal Carolina University, and placed in cold storage.

4.2. X-ray Computed Tomography (CT) Scanning

SBL2 was scanned by a GE Optima 64-slice X-ray CT scanner at Conway Medical

Center in Conway, SC at 0.625 mm vertical resolution, resulting in ~3000 CT images. A three-dimensional CT image of the core, reconstructed using ImageJ, reveals high density laminations interpreted as extreme flood deposits. Extraction of CT number data was automated using a code developed in MATLAB that calculates the 90% cumulative frequency value (CT90) for each core slice by constructing cumulative frequency curves for each slice and selecting the CT value that falls at 90%. CT90 was used as it isolates variations in the dense tail of CT number distributions, which is necessary for paleoflood identification. Sharp spikes in CT90 were interpreted as flood event layers, while low CT90 may indicate normal river stage or drought conditions.

4.3. Laboratory Methods

SBL2 was split in dark-room conditions. One half of the split was sub-sampled for various dating techniques and the other was used for grain-size analysis.

4.3.2. Grain-size Analysis

Grain-size analysis was performed at 0.5 cm resolution using a Cilas 1190 laser diffraction particle-size analyzer. Before measurement, small sub-samples (<1 g) were transferred to 40 ml beakers and reacted with 5 ml of 30% hydrogen peroxide (H2O2) for

24-48 hours to digest organic material. Samples were then heated for 2-3 minutes to remove excess H2O2 and 2 ml of sodium hexametaphosphate (NaPO3)6 was subsequently added as a dispersant to prevent flocculation. Representative sub-samples were obtained following a “wet-splitting” method. This method is performed as follows:

1. Pour a wet sediment sample from a 40 ml beaker into a 600 ml beaker and use a deionized water bottle to ensure all sediment is rinsed into the larger beaker. 2. Add deionized water to the 600 ml beaker until it reaches the 100 ml line. 3. Arrange three empty 40 ml beakers in a row on the lab bench. 4. Begin swirling the 600 ml beaker in a circular motion to get sediments well mixed and continue to do so to keep them in solution. While swirling the larger beaker take pipette draws of equal volume from roughly two-thirds depth in the solution and transfer them into pre-arranged 40 ml beakers in sequential order (begin with the first 40 ml beaker, then the second, then the third, back to the first, and so on). *Make sure to continue swirling the 600 ml beaker the entire time you take pipette draws. A well-mixed sediment solution is key to obtaining a representative sub-sample. 5. Continue swirling the sediments in the beaker and taking pipette draws of equal volume into the 40 ml beakers until the 600 ml beaker is empty. 6. The original grain-size sub-sample is now split into a volume that achieves the correct obscuration value needed for proper measurement using the laser particle size analyzer, CILAS 1190, while still being representative of the true range of grain-sizes in the sediment mixture.

“Wet-splitting” overcomes the difficulty of obtaining a representative sub-sample in multimodal sediment mixtures of mud and sand. This is especially important for studies relying on grain size to index paleoflood in oxbow lakes that sometimes receive a small amount of coarse sediment during extreme flood events.

4.3.3. Radiocarbon Dating

Accelerator mass spectrometry 14C dating was carried out on plant macrofossils to constrain the age of flood layers within the deeper part of the core (> 98 cm). Macrofossils were carefully selected from various depths (Table 1) and cleaned of sediment using an ultrasound bath. Afterward the macrofossils were stored in 10% HCL in glass vials to prevent fungal growth and sent to the National Ocean Sciences Accelerator Mass

Spectrometry facility at Woods Hole Oceanographic Institute for dating measurements. 14C ages were calibrated with the IntCal 13 curve (Reimer et al., 2013) using OxCal online

(Bronk Ramsey, 2009).

4.3.4. 210Pb and 137Cs Dating

Sub-samples (2 cm thick) were taken every 5 cm through the upper part of SBL2

(0-100 cm depth) for 210Pb and 137Cs dating (Table 2). A Canberra Broad Energy

Germanium Detector (BEGe; model no. BE3830p-DET) in the Marine Science Center at

Northeastern University was used to measure 210Pb and 137Cs activities. The constant rate of supply (CRS) model after Appleby (2001) was used to calculate ages and standard error based on the 210Pb profile. Measurements of Cs137 activity were used to identify a peak dated as the 1962 peak atmospheric nuclear bomb testing horizon.

4.3.5. Optically Stimulated Luminescence (OSL)

Three 5 cm thick sub-samples (Table 2) were taken in a dark room laboratory lit by amber and red LED lights at Coastal Carolina University. The outer 2 cm of each sediment sample was separated, dried, and used for dose rate measurements following determination of water content. The remaining sediment was processed to remove organics and carbonates and extract fine silt (4-15 µm) particles following Mauz et al. (2002). Fine-silt samples were dried, transferred to sample bags, stored in light-tight film tubes, and shipped to the OSL dating facility at the University of Salzburg in Austria for luminescence measurement.

4.4. Flood Identification and Paleodischarge Estimate

Bayesian end-member modelling analysis (Yu et al., 2016) was performed to identify a coarse EM of the grain-size data used as a proxy of extreme flood. The coarsest

EM, denoted as EM1, was used for this purpose. EM1was linearly detrended and z-scored between section breaks identified by change point analysis (CPA; Taylor, 2005) to normalize the data. Natural geomorphic changes, damming, and construction of nearby roads in the floodplain of the river alter sediment availability and lithological properties of sediments delivered to the oxbow lake through time. Therefore, normalization is necessary at change points so all sections throughout the core are comparable. Z-scores for each section were computed as

푥 −푦 푧 = 푖 푖 푠 where, x is the detrended EM1 score, y is the section detrended EM1 score average, and s is the section detrended EM1 score standard deviation. Flood events were identified

32 objectively as detrended EM1 z-scores that exceed the mean by a threshold of 1.2 times the standard deviation. The basis of this threshold is to lead to identification of only major flood events of the PDR. A linear relationship between EM1 z-scores and measured discharge of 8 moderate to major flood events (USGS gauge: 02131000) was established for the last ~80 years and applied to the older part of the core to estimate discharge of paleofloods. Additionally, the CT90 data were detrended by a 101-point moving minimum.

Flood events are identified where detrended CT90 exceeds a 71-point moving mean by a one standard deviation threshold of CT90.

4.5. Age-Depth Modelling

An age-depth model was created for SBL2 in the program R using Bacon v2.3.3

(Blaauw and Christen, 2011), an age-depth modelling program that uses Bayesian statistics to develop accumulation histories through combining various dates with prior information.

The initial information to feed the model included the core top age (AD 2018), 210Pb dates,

137Cs peak, OSL ages, and 14C data. Only one 14C sample (194 cm) was able to produce a meaningful calibrated age, and all other samples were either modern or unable to provide a useful calibration age. Within the age-depth model’s uncertainty ranges, depths previously identified as flood layers were assigned a flood year that corresponded to an extreme peak discharge in the instrumental data. The event-discharge correlation that maximized the linear relationship between z-score and peak discharge was chosen for layers that had more than one flood year possibility. After assigning flood years to specific depths, Bacon was re-run with the addition of these historical tie points, which helped decrease the model’s uncertainty.

Several parameters in Bacon were changed from their default setting to improve age-depth model output. The first parameter acc.mean is the average sediment accumulation rate for the core. The parameter thick was set to choose the section thickness for the model and setting it to a thinner thickness results in more sections and a smoother model.

4.6. Flood Frequency Analysis

After flood identification, the number of flood events occurring in a 31-year window were counted. This multidecadal interval is important because decisions on project sizing and policies governing water-project operation are often based on an analysis of 10 to 30 years of data under the assumption that the processes are stationary (National

Research Council, 1998). Pearson’s correlation function was run in MATLAB to assess the relationship between flood frequency and AMO using Kaplan SST V2 data

(NOAA/OAR/ESRL PSD, 2019, Boulder, Colorado, USA; Gray et al., 2004) and PDO indices (Shen et al., 2006).

5. Results

5.1. Core Sedimentology

SBL2 is primarily composed of organic rich mud with bands of less organic sandy silt that are interpreted as flood layers (Figure 11). The upper portion of the core (0-40 cm) is largely comprised of organic-rich mud with a 1 cm thick band of organic material at 40 cm marking a sharp contact between organic rich sediments and less organic rich sediments. From 40-100 cm sediments seem to remain homogenous in color and texture.

The deeper part of the core (>105 cm) exhibits variations in color, texture and organic material. At a few depth intervals within (114-115 cm, 137-139 cm, 144-145 cm, 175-177 cm), grayish-brown mineral rich laminations (~0.5 cm thick) are overlain by finer yellow to pale-yellow sediments, which is characteristic of a couplet-style stratification (Wang et al., 2008). During the peak pulse of a flood, the coarsest mineral rich sediments are deposited first, as flood flows diminish, finer oxidized sediments deposit on top. Additional organic layers existing at 174 cm and 184-186 cm, most likely indicate times of low lake level.

Figure 11: SBL2 core photograph.

5.2 Core Chronology

According to the age-depth model produced by Bacon, SBL2 brackets a time period of AD 1832-2017 (Figure 12). When running Bacon, approximate sediment accretion rates

(acc.mean) of 1 year/cm were initially used, producing an age-depth model that seemed heavily weighted by the rather uncertain bottom 14C date. Since the OSL ages suggested a faster sediment accretion rate towards the bottom of the core the acc.mean was changed to

0.75 year/cm, resulting in an age-model that adhered more closely to the OSL ages.

Additionally, changing the section thickness (thick) in Bacon from the default of 5 cm to 2 cm resulted in a smoother model with less uncertainty. After running the age-depth model, identifying flood events at specific depths, and adding these events as tie points, re-running

Bacon produced a more “step-like” model above 40 cm. This is expected as floods are the primary force driving sedimentation in oxbow lakes and changes in the frequency and intensity of flooding probably result in episodic overbank deposition in floodplain environments (Shen et al., 2015).

137Cs peak

Figure 12: The age-depth model for SBL2 shows the median age probability (black line) and 1σ confidence intervals (bracketed by red-dotted lines), with 1σ error bars on individual chronological controls.

5.3 Grain-size and CT data EM1 was used to define the coarse component of the grain-size data used to trace major paleofloods of the PDR (Figure 13). This EM looks very similar to a raw grain-size distribution of an extreme flood layer (Figure 14). The coarse “shoulder” of EM1, representing deposits of the most extreme flood events, consists of coarse silts and very- fine sands of grain sizes 31 – 90 µm. The finer mode of EM1 is dominated by medium silts

(10 – 20 µm) interpreted as deposition during the waning stages of floods as discharge decreases. The finer mode EM3 resembles grain-size distributions of background sedimentation (Figure 15). Sudden change points at 20 cm and 66 cm in raw EM1 score trace the construction of a road built in the floodplain near SBL2 between AD 1986 and

AD 1987, and the beginning of upstream dam construction in NC in the early 20th century

(Figure 16A), respectively.

Linearly detrending and z-scoring EM1 score between change points helped to normalize the data and made different sections comparable (Figure 16A, 16B). This paleoflood record was able to identify 21 paleoflood events that include all 5 major flood events during the instrumental record period. One flood layer identified in the z-scored

EM1 data (34.5 cm) was unable to be tied to a major flood within the instrumental record.

The CT data was able to cross-validate some of the major flood events identified in the z- scored EM1 data, however the CT method identified more events that are not identified in the grain-size data (Figure 16C).

Figure 13: End member modelling outcomes identified three dominant end members in SBL2.

3.5

2.5

% Volume % 1.5

0.5

0 0.01 0.1 1 10 100 1000 Grain-size (µm)

Figure 14: Grain-size distribution of an extreme paleoflood deposit (115 cm).

5 4.5 4

3.5 3 2.5 2

Volume % 1.5 1 0.5 0 0.01 0.1 1 10 100 1000 Grain-size (µm)

Figure 15. Grain-size distribution of background sediments (87 cm) deposited during non- flooding conditions.

A. B. C.

Figure 16. CT-scan image, EM1 score, z-scored EM1, and CT number for SBL2 reveal coarse mineral-rich layers used as tracers of extreme paleofloods. A. The coarsest EM (EM1 score) has change points (gray dotted lines) at 20 cm and 66 cm that may correspond to a dirt road construction in the floodplain and the beginning of dam construction, respectively. B. Z-scores that exceed the extreme flood threshold (gray dotted line) have been assigned to flood years that fall within the age-depth models 1σ uncertainty range. C. CT data confirm flood events identified in z-scored EM1 data, however some events are overcounted.

5.4 Paleodischarge Estimates A linear model was developed to relate flood bed coarseness to instrumental peak

3 -1 annual discharge (Qmax [m s ]):

Qmax = 3821z – 2528.9 where z is detrended EM1 z-score. The model is significant (p<0.05) and the R-squared value of 0.746 indicates that the model explains about 75% of the variability in the response

(Figure 17). Using this relationship, peak discharges were estimated for every flood event layer identified with the largest on record occurring in approximately AD 1876 (Figure

18). The most intense floods occur between AD 1870-1900. It seems that the most extreme

3 floods (Qmax>5000 m /s) have decreased after AD 1945 (Figure 18).

Figure 17. Scatterplot and linear regression with 1σ prediction intervals relating flood layer z-score to peak annual discharge measurements of the Pee Dee River at the USGS gage station at Pee Dee, SC.

Figure 18. Peak annual discharge estimates (AD 1842 – 1945) with 1σ prediction intervals of paleoflood layers. Paleoflood discharges were estimated using the linear relationship between flood-bed z-score and instrumental peak annual discharge measurements. Peak annual discharges from AD 1945 to AD 2018 are instrumental peak annual discharges. The red bar at the top shows a period of more intense floods, and the blue bar shows a period of decreasing flood magnitude.

5.5 Climate Modes and Flood Frequency Flooding frequency of the PDR exhibits multidecadal variability. Flood frequency is high from AD 1846-1866 and AD 1921-1949 and low from AD 1869-1913 and AD

1953-1972. Correlation between flood frequency and PDO index is positive (Pearson’s

R=0.39) and significant (p<0.001) (Figure 19), while the correlation between AMO and flood frequency is very weak (0.14) and insignificant (Figure 20). Although AMO index and flood frequency are not correlated, three of the largest known hurricane-induced floods

(AD 1928, AD 1945, AD 2018) all occurred when the AMO was in its positive phase.

There is a positive relationship between ENSO index and major flood occurrence on the

PDR as every winter/spring flood during the historical period (AD 1979, AD 1983, AD

2003) has occurred during a positive ENSO index (Figure 21).

Figure 19. PDO index and PDR flood frequency (31-year window) exhibit a positive correlation that is statistically significant (Pearson’s R=0.39, p<0.001).

Figure 20. AMO index and PDR flood frequency exhibit weakly positive correlation that is not statistically significant (Pearson’s R=0.14, p>0.05).

Figure 21. Correlation between major winter/spring flood events (black stars) of the PDR with Multivariate ENSO Index Version 2 calculated using JRA-55 reanalysis (Kobayashi et al., 2015). Figure adapted from NOAA ESRL (2019, https://www.esrl.noaa.gov/psd/enso/mei/).

6. Discussion

6.1. Paleoflood Detection Methods

The results of paleoflood identification indicate that the z-scored coarse EM method of Toonen et al. (2015) is more accurate than the CT detection method. Nevertheless, the

CT data helps to confirm the existence of flood layers. The CT method missed some events, however, and overcounted flood events that were not identified by the grain-size data. The grain-size data is also useful due to its ability to reconstruct paleoflood discharge.

Magnitude of a given flood event is difficult to infer from the CT data because sediment bulk density does not have a direct relationship with flow speed or discharge, while grain size does (Hjulström, 1935). CT data still holds potential for paleoflood research as demonstrated by Støren et al. (2010).

6.2. Climatic and Anthropogenic Controls on Flood Frequency

Flood frequency of the PDR shows a multidecadal trend that is linked to variability of the PDO. PDO controls the frequency of wintertime floods which dominate the PDR flood record, while AMO has an influence on floods caused by hurricanes. For this reason, the following discussion on the PDR flood frequency and its relation to climate modes has been divided into two sections focusing on the different flood seasons.

6.2.1. Summer & Fall Floods

The fact that flood frequency and AMO do not show a significant correlation is not surprising. Enfield et al. (2001) showed that over the PDR catchment there are very weak to zero correlation between AMO phase and precipitation. Most of the Mississippi River drainage basin shows negative correlations between AMO and precipitation, while in southern Florida, AMO and precipitation exhibit a positive correlation (Enfield et al.,

2001). This indicates that the PDR basin is positioned in a transition zone between increased and decreased precipitation caused by positive AMO phase.

Although AMO and flood frequency are not correlated, the five most extreme floods of the PDR occurring in AD 1878, AD 1896, AD 1928, AD 1945, and AD 2018 were unanimously caused by major hurricanes during AMO warm phases (Figure 22). This may be due to increased hurricane landfalls on the east coast, which increase during AMO warm phases (Goldenberg et al., 2001; Zhang & Delworth, 2006). The lack of correlation between AMO and flood frequency is probably because hurricane caused floods are relatively rare along the PDR compared to wintertime floods, and do not occur often enough to impact the long-term frequency of flooding of the PDR.

Figure 22. Correlation between the most extreme floods caused by hurricanes (black stars) and AMO index. Adapted from Trenberth and Zhang (2019).

6.2.2. Winter & Spring Floods

Although, the most extreme flood events of the PDR were caused by hurricanes, most major floods in the instrumental record occur in the winter and spring months. An increased frequency of flooding along the PDR arises from the combined interaction of positive PDO phases and El Niño, resulting in an amplified climate response to El Niño

(Gershunov & Barnett, 1998; Enfield et al., 2001; Fuentes-Franco et al., 2016).

El Niño events are shown to have a strong relationship with colder wetter winters in the southeast U.S. (Ropelewski & Halpert, 1987; Kurtzman & Scanlon, 2007; Maleski

& Martinez, 2018). Composite anomalies of 500 mb geopotential heights during winter El

Niño events are negative over the PDR catchment (Figure 23), creating a trough in pressure and temperature that is conducive to enhanced extratropical cyclone and frontal system activity. These factors result in a persistent extended Pacific jet stream and an amplified storm track across the southern tier of the US (Figure 24). As a result, the southeast experiences positive wintertime precipitation anomalies (Figure 25) which increase soil moisture. Additionally, there is a higher probability for extreme precipitation to occur

(Figure 26) and an increased precipitation rate (Figure 27). This indicates that during El

Niño winters extreme precipitation has a higher probability of falling on saturated soils at an increased rate, thus increasing the likelihood of extreme flooding events along the PDR.

These extreme precipitation events are cause by wintertime extratropical cyclones and their associated fronts. During El Niño events there are negative 500 mb geopotential height anomalies over the PDR catchment. These lower pressure conditions create more favorable conditions for extreme winter storms and cold weather.

Positive PDO phases have the propensity to influence an enhanced frequency of El

Niño events (Verdon & Franks, 2006). Since El Niño events lead to flood favorable conditions in the PDR catchment, more frequent El Niño events leads to more frequent flooding, explaining the positive correlation between PDO and flood frequency. The positive PDO correlation with flood frequency is an exciting finding, supporting claims of strong teleconnections between PDO and ENSO during winter (Gershunov & Barnett,

1998; Gershunov & Cayan 2003; Pavia et al. 2006). This paper confirms that when PDO and ENSO coincide in their positive phases a more robust pattern of wetter conditions exists in the southern part of the United States (Gershunov & Barnett, 1998; Enfield et al.,

2001; Fuentes-Franco et al., 2016).

Periods of increased PDR flood frequency coincide with periods of increased spring rainfall reconstructed using SC tree-ring proxies (Stahle & Cleveland, 1992). Although most of the flooding on the PDR occurs from December-April, the correlation between flood frequency and spring rainfall shows that general wetness and flooding are connected in SC. General wetness increases soil moisture and adds strength to the argument that positive precipitation anomalies during El Niño events leads to flood favorable conditions.

An inaccurate representation of flooding frequency by the paleoflood reconstruction in this paper after AD 1980 is due to land use change near SBL. Sediment grain size shifts coarser after AD 1985, and never seem to return to background levels.

There is evidence of a dirt road being built near SBL by aerial photographs on google earth after AD 1986 (Figure 28). Floods occurring after the construction of this road probably transport sands from the road and into SBL limiting the reliability of flood frequency reconstruction after AD 1987.

Figure 23. Composite anomaly of 500mb geopotential height over North America during El Niño events. Trough in geopotential height is evident across the southern tier of the US. Image courtesy of NOAA’s Earth System Research Laboratory.

Figure 24. Effects on general temperature, wetness, and storminess of North America during El Niño winters.

Figure 25. Composite precipitation anomalies over the US during El Niño winters. Positive precipitation anomalies exist over the PDR catchment. Image courtesy of NOAA’s Earth System Research Laboratory.

Figure 26: Chance of extreme precipitation over the US during El Niño winters (Left: November, December, January; Right: December, January, February). Extreme precipitation is defined as being in the wettest 20% of the 100-year record. Wet extremes are +90% increase in risk for the majority of the PDR catchment during El Niño events. Image courtesy of NOAA’s Earth System Research Laboratory.

Figure 27: Composite anomaly of precipitation rate (mm/day) over the US during El Niño winters. Increased precipitation rate is observed over the PDR catchment area in NC and SC. Image courtesy of NOAA’s Earth System Research Laboratory.

Figure 28: Google earth maps showing a dirt road constructed between AD 1986 and AD 1987 near SBL2. The white arrow on the right photo points to the road.

6.3 Change of Peak Annual Discharge

A striking observation about peak annual discharge change in this paleoflood record is that many events prior to the instrumental period produced discharges that are at least the magnitude of the AD 1945 event or larger. The most intense paleofloods of the PDR occur within the time period of AD 1870-1900. An analysis of tropical storm and hurricane tracks that pass through the PDR catchment from AD 1851-2018 shows that AD 1870-

1900 corresponds to a period of heightened storm activity (Figure 29), which likely caused intensified flooding of the PDR. Storm tracks of tropical storms/hurricanes in AD 1878

(Figure 30), AD 1896 (Figure 31), AD 1928 (Figure 32), and AD 1945 (Figure 33) all travel directly through the PDR basin.

Additionally, Trimble (2008) showed that from AD 1860-1920 most of the PDR basin was experiencing peak levels of erosive land use for farming of cotton and tobacco

(Figure 34). Clear cutting of forests for agriculture, likely resulted in a decrease in natural land cover and a faster rate of runoff into streams of the PDR basin. The most intense flood magnitudes on record, therefore, were probably caused by the combined effects of increased tropical cyclone activity and land-use change.

As hypothesized, there has been a general decrease in the peak annual discharge of major floods in the PDR since dam construction began in the early 20th century. This finding is not only found in the PDR. A regional signal of decrease in peak annual discharge of dammed rivers is apparent. Peak annual discharge data of the Saluda River at Chappells,

SC (Figure 35), the Congaree River at Columbia, SC (Figure 36), and the Savannah River at Burtons Ferry bridge near Millhaven, GA (Figure 37) all show decreases of peak annual discharge after AD 1940. When viewing peak annual discharge data of free-flowing rivers

54 in the SEACP, such as the Edisto River near Givhans, SC (Figure 38) or the Waccamaw

River near Longs, SC (Figure 39), this decrease after AD 1940 is nonexistent. Instead, there is an apparent increase in peak annual discharge of the most extreme events. Together, these discharge records suggest that river management probably has reduced the risk for extreme flooding in much of the SEACP

Although recent flooding of the PDR is not unprecedented in the context of the long-term record generated by this study, flood risk of rivers in the SEACP still remains heightened, especially for free-flowing rivers. Extreme events like the SC rainfall event of

2015 (Figure 40) are projected to increase in frequency and severity through the 21st century (Wuebbles et al., 2017). Currently increasing development in the SEACP, leads to increases in land clearing and construction of impervious surfaces that have been shown to heighten flood magnitudes of the past. Not only does increasing development elevate extreme flood risk, but it results in population growth leading to higher costs and possibly more deaths resulting from inland flooding.

Figure 29. Number of tropical storms and hurricanes affecting the PDR basin from AD 1851-2018. 1878

Figure 30. Storm track of a hurricane in AD 1878. Data obtained from NOAA Office of Coastal Management.

1896

Figure 31. Storm track of a hurricane in AD 1896. Data obtained from NOAA Office of Coastal Management. 1928

Figure 32. Storm Track of a hurricane in AD 1928. Data obtained from NOAA Office of Coastal Management.

1945

Figure 33. Storm track of a Hurricane in AD 1945 that caused the largest peak annual discharge in the instrumental record on September 22nd, 1945. Data obtained from NOAA Office of Coastal Management.

Figure 34. Level of erosive land use in the southern Piedmont from AD 1700-1967. Figure adapted from Trimble (2008).

Figure 35. Peak annual discharge of the Saluda River at Chappels, SC.

Figure 36. Peak annaual discharge of the Congaree River at Columbia, SC.

Figure 37. Peak annual discharge of the Savannah River at Burtons Ferry bridge near Mill Haven, GA.

Figure 38. Peak annual discharge of the Edisto river near Givhans, SC.

Figure 39. Peak annual discharge of the Waccamaw River near Longs, SC.

Figure 40. Precipitation map of the historic SC rainfall event of 2015. Adapted from CISA (2016).

7. Conclusions

Grain-size analysis and EM modelling combined with multiple dating techniques, provides an almost two-century long continuous event-scale record of paleofloods with well resolved peak annual discharge estimates. Paleoflood frequency of the PDR exhibits a multidecadal trend that is positively correlated with PDO index. The most extreme events in this record were caused by the interaction of heightened tropical cyclone activity (during

AMO warm phases) with land-use change in the PDR catchment. Additionally, dams have greatly reduced flood intensity of many rivers in the SEACP since the mid-20th century.

This study builds upon evidence in the literature that high resolution paleoflood data can be obtained from oxbow lake sediments. Prior to this paleoflood reconstruction, relatively little success in resolving paleoflood discharges has been achieved in the SEACP.

This research opens many doors for future paleoflood research in the SEACP where a recent increase in costly and deadly floods is occurring.

Tables Table 1. List of macrofossils used for radiocarbon dating.

Age Err Date Submitter F 13C Age (14C (14C 13C Reported Identification Type Process Modern Fm Err Corr yr) yr) 13C Source 14C (OC) Organic - - 5/30/2018 SBL2_C1A Tree bark Carbon 0.9974 0.0021 * 20 15 24.75 Measured 10.79 10 small (OC) Organic - - 6/5/2018 SBL2_C1B leaves Carbon 0.9846 0.0021 * 125 15 28.32 Measured 23.48 (OC) Organic - 5/30/2018 SBL2_C2 3 leaves Carbon 0.9994 0.0023 * 5 20 30.68 Measured -8.81 (OC) Organic - 5/30/2018 SBL2_C3 1 leaf Carbon 1.0024 0.0028 * >Modern 28.66 Measured -5.81 (OC) Organic - 5/30/2018 SBL2_C4 1 acorn Carbon 1.0003 0.0021 * >Modern 27.69 Measured -7.89

Table 2. Dating samples used in this study and their corresponding depths.

Core: SBL2 210Pb Dating Sub-samples 14C Dating Sub-samples OSL Dating Sub-samples Sample ID Depth (cm) Sample ID Depth (cm) Sample ID Depth (cm) SBL02_L01 0-2 cm SBL2_C1A 194-195 cm SBL2_OSL1 180-185 cm SBL02_L02 5-7 cm SBL2_C1B 194-195 cm SBL2_OSL2 120-125 cm SBL02_L03 10-12 cm SBL2_C2 191-192 cm SBL2_OSL3 80-85 cm SBL02_L04 15-17 cm SBL2_C3 143-144 cm SBL02_L05 20-22 cm SBL2_C4 98-99 cm SBL02_L06 25-27 cm SBL02_L07 30-32 cm SBL02_L08 35-37 cm SBL02_L09 40-42 cm SBL02_L10 45-47 cm SBL02_L11 50-52 cm SBL02_L12 55-57 cm SBL02_L13 60-62 cm SBL02_L14 65-67 cm SBL02_L15 70-72 cm SBL02_L16 75-77 cm SBL02_L17 78-80 cm SBL02_L18 85-87 cm SBL02_L19 90-92 cm SBL02_L20 95-97 cm

References Baker VR. 1987. Paleoflood hydrology and extraordinary flood events. Journal of Hydrology 96: 79-99.

Baker VR. 2008. Paleoflood hydrology: origin, progress, prospects. Geomorphology 101: 1-13.

Blaauw M, Christen JA. 2011. Flexible paleoclimate age–depth models using an autoregressive gamma process. Bayesian Analysis 6:457–474.

Blott SJ, Pye K. 2006. Particle size distribution analysis of sand-sized particles by laser diffraction: an experimental investigation of instrument sensitivity and the effects of particle shape. Sedimentology 53(3):671-685. doi: 10.1111/j.1365- 3091.2006.00786.x

Bronk Ramsey, C. (2009). Bayesian analysis of radiocarbon dates. Radiocarbon 51(1): 337-360.

Chow VT, Maidment DR, Mays LW. 1988. Applied Hydrology. New York (NY): McGraw-Hill Inc. 598 p.

CISA. 2016: The South Carolina Floods of October 2015. Carolinas Integrated Sciences & Assessments (CISA), Columbia, SC, 4 pp.

Cook CW. 1936. Geology of the coastal plain of South Carolina. Reston (VA): US Department of the Interior, United States Geologic Survey. Bulletin 867.

Dankers R, Arnell NW, Clark DB, Falloon PD, Fekete BM, Gosling SN, Heinke J, Kim H, Masaki Y, Satoh Y, Stacke T, Wada Y, Wisser D. 2014. First look at changes in flood hazard in the inter-sectoral impact model intercomparison project ensemble. PNAS 111(9):3257-3261.

Delworth T, Manabe S, Stouffer RJ. 1993. Interdecadal Variations of the Thermohaline Circulation in a Coupled Ocean-Atmosphere Model. Journal of Climate 6(11):1993-2011.

Delworth TL, Manabe S, Stouffer RJ. 1997. Multidecadal climate variability in the Greenland Sea and surrounding regions: A coupled model simulation. Geophysical Research Letters 24(3):257-260. doi: doi:10.1029/96GL03927

Delworth TL, Mann, ME. 2000. Observed and simulated multidecadal variability in the Northern Hemisphere. Climate Dynamics 16(9):661-676. doi: 10.1007/s003820000075

Deser, C, Trenberth, K & National Center for Atmospheric Research Staff (Eds). Last modified 06 Jan 2016. "The Climate Data Guide: Pacific Decadal Oscillation (PDO): Definition and Indices." Retrieved from https://climatedataguide.ucar.edu/climate-data/pacific-decadal- oscillation-pdo-definition-and-indices.

Dietze E, Hartmann K, Diekmann B, IJmker J, Lehmkuhl F, Opitz S, Stauch G, Wünnemann B, Borchers A. 2012. An end-member algorithm for deciphering modern detrital processes from lake sediments of Lake Donggi Cona, NE Tibetan Plateau, China. Sedimentary Geology 243-244:169-180.

Dietze E, Maussion F, Ahlborn M, Diekmann B, Hartmann K, Henkel K, Kasper T, Lockout G, Opitz S, Haberzettl T. 2013. Sediment transport processes across the Tibetan Plateau inferred from robust grain size end-members in lake sediments. Climate of the Past 9(4):4855-4892.

Dietze M, Dietze E. 2016. EMMAgeo: End-Member Modelling of Grain-Size Data. R package version 0.9.6.

Donat MG, Lowry AL, Alexander LV, O’Gorman PA, Maher N. 2016. More extreme precipitation in the world’s dry and wet regions. Nature Climate Change 6: 508- 513.

Duliu OG. 1999. Computer axial tomography in geosciences: an overview. Earth-Science Reviews 48(4):265-281.

Ely LL. 1997. Response of extreme floods in the southwestern United States to climatic variations in the late Holocene. Geomorphology 19(3):175-201.

Ely LL, Baker VR. 1985. Reconstructing paleoflood hydrology with slackwater deposits- Verde River Arizona Physical Geography 6(2):103-126.

Enfield DB, Mestas-Nuñez AM, Trimble PJ. 2001. The Atlantic Multidecadal Oscillation and its relation to rainfall and river flows in the continental U.S. Geophysical Research Letters 28(10):2077-2080. doi: doi:10.1029/2000GL012745

Fisher MM, Brenner M, Reddy KR. 1992. A simple, inexpensive piston corer for collecting undisturbed sediment/water interface profiles. Journal of Paleolimnology 7:157.

Fisher P, Aumann C, Chia K, O’Halloran N, Chandra S. 2017. Adequacy of laser diffraction for soil particle size analysis. PLoS ONE 12(5):e0176510.

Folland CK, Parker DE, Kates FE. 1984. Worldwide marine te,perature fluctuations 1856- 1981. Nature 310: 670-673.

Folland CK, Palmer TN, Parker DE. 1986. Sahel rainfall and worldwide sea temperatures, 1901-85. Nature 320: 602-607.

Frazier M, Shelton J, Hall T, Park W, Streubel D, Hom V, et al. 2016. The Historic South Carolina Floods of October 1-5, 2015. Silver Spring (MD): US Department of Commerce, National Oceanic and Atmospheric Administration, National Weather Service.

Gershunov A, Barnett TP. 1998. Interdecadal Modulation of ENSO Teleconnections. Bulletin, American Meteorological Society 79: 2715-2726.

Gershunov A, Cayan DR. 2003. Heavy Daily Precipitation Frequency over the Contiguous United States: Sources of Climatic Variability and Seasonal Predictability. Journal of Climate 16: 2752-2765.

Glur L, Wirth SB, Büntgen U, Gilli A, Haung GH, Schär C, Beer J, Anselmetti FS. 2013. Frequent floods in the European Alps coincide with cooler periods of the past 2500 years. Scientific Reports 3:2770.

Goldenberg SB, Landsea CW, Mestas-Nuñez AM, Gray WM. 2001. The Recent Increase in Atlantic Hurricane Activity: Causes and Implications. Science 293(5529):474- 479. doi: 10.1126/science.1060040

Goman M, Leigh D. 2004. Wet early to middle Holocene conditions on the upper Coastal Plain of North Carolina, USA. Quaternary Research, 61(3), 256-264. Gray ST, Graumlich LJ, Betancourt JL, Pederson GT. 2004. A tree-ring based reconstruction of the Atlantic Multidecadal Oscillation since 1567 A.D. Geophysical Research Letters 31(12).

Hirschboeck KK. 1991. USGS National water summary 1988-89 – Hydrology of floods and droughts: Climate and floods. Reston (VA): US Department of the Interior, United States Geologic Survey. Open-file reports, Water-Supply Paper 2375:67- 88.

Hjulström, F. 1935. Studies of the morphological activity of rivers as illustrated by the River Fyris. Geological Institute Upsala Bulletin 25: 221-527.

Kerr RA. 2000. A North Atlantic Climate Pacemaker for the Centuries. Science, 288(5473), 1984-1985. doi: 10.1126/science.288.5473.1984

Knox JC. 1993. Large increases in flood magnitude in response to modest changes in climate. Nature 361:430. doi: 10.1038/361430a0

Knox JC. 2000. Sensitivity of modern and Holocene floods to climate change. Quaternary Science Reviews 19(1):439-457.

Kobayashi S, Ota Y, Harada Y, Ebita A, Moriya M, Onoda H, Onogi K, Kamahori H, Kobayashi C, Endo H, Miyaoka K, Takahashi K, 2015. The JRA-55 Reanalysis: general specifications and basic characteristics. J. Meteor. Soc. Jpn. 93, 5-48.

Kochel RC, Baker VR. 1982. Paleoflood hydrology. Science 215:353- 361.

Konert M, Vandenberghe J. 1997. Comparison of laser grain size analysis with pipette and sieve analysis: A solution for the underestimation of the clay fraction. Sedimentology 44(3):523-535.

Kurtzman D, Scanlon BR. 2007. El Niño–Southern Oscillation and Pacific Decadal Oscillation impacts on precipitation in the southern and central United States: Evaluation of spatial distribution and predictions. Water Resources Research 43(10).

Leigh DS, Srivastava P, Brook GA. 2004. Late Pleistocene braided rivers of the Atlantic Coastal Plain, USA. Quaternary Science Reviews 23(1-2):65-84.

Leigh DS. 2008. Late Quaternary climates and river channels of the Atlantic Coastal Plain, Southeastern USA. Geomorphology 101(1-2): 90-108.

Leigh DS. 2018. Vertical accretion sand proxies of gaged floods along the upper Little Tennessee River, Blue Ridge Mountains, USA. Sedimentary Geology 364:342- 350.

Maleski JJ, Martinez CJ. 2018. Coupled impacts of ENSO, AMO, and PDO on temperature and precipitation in the Alabama–Coosa–Tallapoosa and Apalachicola– Chattahoochee–Flint river basins. International Journal of Climatology 38(S1):e717-e728.

Mann ME, Emanuel KA. 2006. Atlantic hurricane trends linked to climate change. Eos, Transactions American Geophysical Union 87(24):233-241. doi: 10.1029/2006eo240001

Mann ME, Steinman, BA, Miller SK. 2014. On forced temperature changes, internal variability, and the AMO. Geophysical Research Letters 41(9):3211-3219.

McCarney-Castle K, Voulgaris G, Kettner AJ. 2010. Analysis of fluvial suspended sediment load contribution through Anthropocene history to the South Atlantic Bight Coastal Zone. The Journal of Geology 118(4):399-416.

McKenzie NJ, Jacquier DJ, Isbell RF, Brown KL. 2004. Australian Soils and Landscapes an Illustrated Compendium. CSIRO Publishing: Collingwood, Victoria.

Messmer M, Gomez-Navarro JJ, Raible CC. 2015. Climatology of Vb cyclones, physical mechanisms and their impact of extreme precipitation over Central Europe. Earth System Dynamics 6: 541-553.

Mizzell H, Simmons J. 2015. South Carolina’s Climate report Card: The influence of the El Niño Southern oscillation Cold and Warm Event Cycles on South Carolina’s Seasonal precipitation. Journal of South Carolina Water Resources 2(1): 3-10.

Morton RA. 2003. An overview of coastal land loss: with emphasis on the southeastern United States. Reston (VA): US Department of the Interior, US Geological Survey. Open-File Report 2003-337.

Munoz SE, Gruley KE, Massie A, Fike DA, Schroeder S, Williams JW. 2015. Cahokia’s emergence and decline coincided with shifts of flood frequency on the Mississippi River. Proceedings of the National Academy of Sciences 112(20):6319-6324.

Munoz SE, Giosan L, Therrell MD, Remo JWF, Shen Z, Sullivan RM, Wiman C, O’Donnell M, Donnelly JP. 2018. Climate variability, river engineering, and unprecedented flood risk along the Mississippi River. Nature 556:95-98.

Musser JW, Watson KM, Painter JA, and Gotvald AJ. 2016. Flood-inundation maps of selected areas affected by the flood of October 2015 in central and coastal South Carolina. Reston (VA): US Department of the Interior, US Geological Survey. Open-File Report 2016-1019. http://dx.doi.org/10.3133/ofr20161019.

National Research Council. 1998. Decade-to-Century-Scale Climate Variability and Change: A Science Strategy. Washington, DC: The National Academic Press. https://doi.org/10.17226/6129.

NOAA National Centers for Environmental Information (NCEI) U.S. Billion-Dollar Weather and Climate Disasters (2018). https://www.ncdc.noaa.gov/billions/.

Orsi TH, Edwards CM, Anderson AL. 1994. X-ray computed tomography: a nondestructive method for quantitative analysis of sediment cores. Journal of Sedimentary Research Section A: Sedimentary Petrology and Processes 64A(3): 690-693.

Paasche Ø, Støren EWN. 2014. How Does Climate Impact Floods? Closing the Knowledge Gap. Eos, Transactions American Geophysical Union 95(28):253-254.

Patton PC, Dibble DS. 1982. Archeologic and geomorphic evidence for the paleohydrologic record of the Pecos River in west Texas. American Journal of Science 282: 97-121.

Pavia, EG, Graef F, Reyes J. 2006. PDO-ENSO Effects in the Climate of Mexico. Journal of Climate 19: 6433-6438.

Philander SGH. 1983. El Niño Southern Oscillation phenomena. Nature, 302:295.

Phillips JD. 1991. Fluvial sediment delivery to a coastal plain estuary in the Atlantic drainage of the United States. Marine Geology 98:121-134.

Reimer PJ, Bard E, Bayliss A, Beck JW, Blackwell PG, Bronk Ramsey C, Grootes PM, Guilderson TP, Haflidason H, Hajdas I, Hatt C, Heaton TJ, Hoffmann DL, Hogg AG, Hughen KA, Kaiser KF, Kromer B, Manning SW, Niu M, Reimer RW, Richards DA, Scott EM, Southon JR, Staff RA, Turney CSM, van der Plicht J. 2013. IntCal13 and Marine13 Radiocarbon Age Calibration Curves 0-50,000 Years cal BP. Radiocarbon, 55(4): 1869-1887.

Ropelewski CF, Halpert MS. 1987. Global and Regional Scale Precipitation Patterns Associated with the El Niño/Southern Oscillation. Monthly Weather Review, 115(8):1606-1626.

Schillereff DN, Chiverrell RC, Macdonald N, Hooke JM. 2016. Hydrological thresholds and basin control over paleoflood records in lakes. Geology 44(1):43-46.

Shen Z, Törnqvist TE, Mauz B, Chamberlain EL, Nijhuis AG, Sandoval L. 2015. Episodic overbank deposition as a dominant mechanism of floodplain and delta-plain aggradation. Geology 43(10): 875-878.

Steinman BA, Mann ME, Miller SK. 2015. Atlantic and Pacific multidecadal oscillations and Northern Hemisphere temperatures. Science 347(6225):988-991.

Støren EN, Dahl SO, Nesje A, Paasche Ø. 2010. Identifying the sedimentary imprint of high-frequency Holocene river floods in lake sediments: development and application of a new method. Quaternary Science Reviews 29(23):3021-3033.

Taylor WA. 2000. Change-Point Analysis: A Powerful New Tool for Detecting Changes. http://www.variation.com/cpa/tech/changepoint.html.

Toonen WHJ, Kleinhans MG, Cohen KM. 2012. Sedimentary architecture of abandoned channel fills. Earth Surface Processes and Landforms 37(4):459-472.

Toonen WHJ, Winkels TG, Cohen KM, Prins MA, Middelkoop H. 2015. Lower Rhine historical flood magnitudes of the last 450years reproduced from grain-size measurements of flood deposits using End Member Modelling. CATENA 130(Supplement C):69-81.

Trimble SW. 2008. Man-Induced soil erosion of the southern piedmont 1700-1970. 2nd ed. Ankeny (IA): Soil and Water Conservation Society.

Trenberth, K, Zhang, R & National Center for Atmospheric Research Staff (Eds). Last modified 10 Jan 2019. "The Climate Data Guide: Atlantic Multi-decadal Oscillation (AMO)." Retrieved from https://climatedataguide.ucar.edu/climate- data/atlantic-multi-decadal-oscillation-amo.

U.S. National Weather Service. (2016). The Historic South Carolina Floods of October 1- 5, 2015.

Van Bebber W. 1891. Die Zugstrassen der barometrischen Minima nachden Bahnenkarten der deutschen Seewarte für den Zeitraum1875–1890, Meteorolgische Zeitschrift 8: 361–366.

Verdon, DC, Franks SW. 2006. Long-term behaviour of ENSO: Interactions with the PDO over the past 400 years inferred from paleoclimate records. Geophysical Research Letters 33, L06712. doi:10.1029/2005GL025052.

Wang L, Leigh DS. 2012. Late-Holocene paleofloods in the Upper Little Tennessee River valley, Southern Blue Ridge Mountains, USA. The Holocene, 22(9), 1061–1066.

Wang Y, Wang Y, Tang L. 2008. On the formation of couplet-style stratifications. International Journal of Sediment Research 23 (1): 85-91.

Wang J, Yang B, Ljungqvist FC, Luterbacher J, Osborn TJ, Briffa KR, Zorita E. 2017. Internal and external forcing of multidecadal Atlantic climate variability over the past 1200 years. Nature Geoscience. doi: 10.1038/ngeo2962

Wachob A, Park AD, Newcome, Jr. R. 2009. South Carolina state water assessment: second edition. Columbia (SC): Land, Water and Conservation Division, SC Department of Natural Resources.

Wilhelm B, Arnaud F, Sabatier P, Crouzet C, Brisset E, Chaumillon E, Disnar JR, Guiter F, Malet E, Reyss JL, Tachikawa K, Bard E, Delannoy JJ. 2012. 1400 years of extreme precipitation patterns over the Mediterranean French Alps and possible forcing mechanisms. Quaternary Research 78(1):1-12.

Wilhelm B, Arnaud F, Sabatier P, Magand O, Chapron E, Courp T, Tachikawa K, Fanget B, Malet E, Pignol C, Bard E, Delannoy JJ. 2013. Paleoflood activity and climate change over the last 1400 years recorded by lake sediments in the north-west European Alps. Journal of Quaternary Science 28(2):189-199.

Withjack EM, Graham SK, Yang CT. 1990. Determination of heterogeneities and miscible displacement characteristics in corefloods by CT scanning: 65th Annual Technical Conference and Exhibition of the Society of Petroleum Engineers, SPE 20490:213- 224.

Wolfe BB, Hall RI, Last WM, Edwards TWD, English MC, Karst-Riddoch TL, Paterson A, Palmini R. 2006. Reconstruction of multi-century flood histories from oxbow lake sediments, Peace-Athabasca Delta, Canada. Hydrological Processes 20(19):4131-4153.

Wuebbles DJ, Fahey DW, Hibbard KA, Dokken DJ, Stewart BC, Maycock TK. 2017. Climate Science Special Report: Fourth National Climate Assessment. Washington, DC: Global Change Research Program. Volume I, Chapter 7,8:470 pp, doi: 10.7930/J0J964J6.

Zhang R, & Delworth TL. 2006. Impact of Atlantic multidecadal oscillations on India/Sahel rainfall and Atlantic hurricanes. Geophysical Research Letters, 33(17). doi:10.1029/2006GL026267