Quantifying Inter-And Supratidal Facies-Belt Migration from Vintage Aerial Photography

Please do not remove this page

Quantifying Inter-and Supratidal Facies-Belt Migration from Vintage Aerial Photography

Wu, Mingyue https://scholarship.miami.edu/discovery/delivery/01UOML_INST:ResearchRepository/12372047100002976?l#13372047090002976

Wu, M. (2020). Quantifying Inter-and Supratidal Facies-Belt Migration from Vintage Aerial Photography [University of Miami]. https://scholarship.miami.edu/discovery/fulldisplay/alma991031524281402976/01UOML_INST:ResearchR epository

Open Downloaded On 2021/09/26 09:39:17 -0400

Please do not remove this page

UNIVERSITY OF MIAMI

QUANTIFYING INTER-AND SUPRATIDAL FACIES-BELT MIGRATION FROM VINTAGE AERIAL PHOTOGRAPHY

Mingyue Wu

A THESIS

Submitted to the Faculty of the University of Miami in partial fulfillment of the requirements for the degree of Master of Science

Coral Gables, Florida

December 2020

UNIVERSITY OF MIAMI

A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science

QUANTIFYING INTER-AND SUPERTIDAL FACIES-BELT MIGRATION FROM VINTAGE AERIAL PHOTOGRAPHY

Mingyue Wu

Approved:

______Sam Purkis, Ph.D. Gregor Eberli, Ph.D. Professor and Chair of Marine Geosciences Professor of Marine Geosciences

______Paul M Harris, Ph.D. Guillermo Prado, Ph.D. Adjunct Professor of Marine Geosciences Dean of the Graduate School

______Miles Frazer, Ph.D. Chevron

WU, MINGYUE (M.S., Marine Geosciences)

Quantifying Inter-and Supratidal Facies-Belt (December 2020) Migration from Vintage Aerial Photography

Abstract of a thesis at the University of Miami.

Thesis supervised by Professor Sam Purkis. No. of pages in text. (85)

Facies-belt dynamics are poorly understood at timescales of decades to centuries because of the lack of quantitative data spanning these time periods. A wealth of vintage- military aerial photography acquired in the years surrounding the Second World War exists, however, and can be paired with modern high-resolution satellite imagery to quantify temporal change over sufficiently long periods to allow meaningful extrapolation to geological timescales. This project focuses on quantifying the dynamics of inter- and supratidal carbonates at two ends of the energy spectrum. For the high energy, I follow

Purkis et al. (2016) and consider the 40-yr. migration of the coastlines of 20 atoll islands situated atop isolated carbonate platforms in the Chagos Archipelago (Indian Ocean). My data set indicates the these fundamental characteristics of coastline behaviors: (i) coastlines facing the prevailing trade winds retreat through time, while those in leeward positions tend to expand; (ii) coastline expansion and retreat are in balance such that total land area of all the considered islands are virtually static over the last 50 years through using commonly accepted +/-3% threshold; (iii) small islands (<20 ha) are substantially more dynamic than large ones. It is worth noting that the stretches of the Diego Garcia coastline in close vicinity to human modification are more likely to suffer erosion than those situated far from human activity, and more likely to erode than those on the uninhabited islands of

Peros Banhos. A comparison between the behavior of a broad portfolio of islands spanning the Indian and Pacific Oceans emphasizes that local factors to be better determinants of coastline dynamics than the rate of eustatic sea-level rise. In aggregate, our data suggest that atoll islands are likely to persist in the face of accelerating sea-level rise, but their high rate of coastline dynamics might plausibly challenge habitability. For low-energy, I examine the tidal flats of Andros Island (Bahamas) – one of the “classic” venues for comparative carbonate sedimentology. The time-separated data assembled for Andros spans 75 years. My results suggest that: (i) the seaward margin of the intertidal deposit remains static in the face of rising sea level, but the deposit’s landward margin retrogrades markedly, broadening the intertidal channelized zone; (ii) in unison to the broadening of the channelized zone, the abundance of laminated (Scytonema) cyanobacterial fabrics within it decreases, accompanied by the lengthening and avulsion of the network of tidal channels that traverse this zone; (iii) the prominence of these changes to the architecture of the flats are surprising, given that the amplitude of sea level rise during the 30 yr. period of observation was only 10 cm. Through the concept of comparative sedimentology, it is anticipated that my observations from the modern might be applied to better recognize the onset of the transgression in ancient carbonate deposits, even for low-amplitude, sub- orbital sea-level cycles.

Acknowledgement

Here, I would thank all the people who helped me during my M.S. study in Miami. Such an unforgettable experience has had a profound influence on my future life.

First of all, I would like to thank my supervisor, Prof. Sam Purkis, who offered me a great chance to study in UM and become one of the members of MGS. During my research experience, he is the first person, who guides me in any fields of doing scientific research from writing codes making figures, forming creative ideas and delivering impressive presentations. He is always there to support me when I need scientific mentorship.

I am also grateful to my committee members - Drs. Paul (Mitch) Harris, Gregor Eberli, and Miles Frazer - for their support and guidance throughout my research. I would also thank Drs. Imelda Johnson and Juan Carlos Laya, for their generous invitation to participate in a fieldtrip to Andros Island with colleagues from TAMU. My gratitude also extends to Dr. Art Gleason, who helped me build hydrodynamic models for the Chagos Archipelago and to Dr. Virginie Duvat who gave me valuable comments for my manuscript which considers the Chagos Archipelago.

I would also give my sincere thanks to my team members (Lisa Tanh, Cecilia Lopez-Gamundi, and Anna Bakker) and all of MGS colleagues and staff. I especially would like to thank my roommates Tianshu Kong and Xiangmin Wang, who accompanied me and encouraged me through the many of problems that I encountered.

Finally, I send my deepest appreciation to my family members for their kindest support through my whole life.

iii

Table of Contents

LIST OF FIGURES ...... vi LIST OF TABLES ...... xi Chapter 1: Introduction ...... 1 Chapter 2: High-Energy Facies Dynamics in the Central Indian Ocean ...... 2 2.1 Background ...... 2 2.2 Regional Setting ...... 7 2.3 Material and Methods ...... 10 2.3.1 Pre-Processing and Georectification of the Vintage Aerial Photographs ...... 10 2.3.2 Pre-Processing of Modern WorldView-2 Satellite Imagery ...... 11 2.3.3 Coastline Extraction and Development of Coastline-Shift Maps ...... 12 2.3.4 Partitioning the Island Coastlines for Timeseries Analysis ...... 15 2.3.5 Sources of Error from Coastline Extraction ...... 17 2.3.6 Hydrodynamic Modelling of Peros Banhos and Diego Garcia ...... 19 2.4 Results ...... 22 2.4.1 Changes Through Time in the Areas of the Peros Banhos and Diego Garcia Atoll Islands ...... 22 2.4.2 Coastline Dynamics Very with Island Shape and Size ...... 31 2.4.3 Spatial Patterns of Island Alteration in Peros Banhos and Diego Garcia ...... 32 2.4.4 Hydrodynamic Setting of Peros Banhos and Diego Garcia ...... 33 2.5 Discussion ...... 37 2.5.1 Attempt for Unraveling Anthropogenic Influence on Island Behavior ...... 38 2.5.2 The Behavior of the Chagos Islands is Consistent with Atoll Islands Globally ...... 40 2.5.3 Rate of Sea Level Rise is a Poor Predictor of Atoll-Island Behavior ...... 42 Chapter 3: Low-Energy Facies Dynamics in Andros Island, Bahamas ...... 45 3.1 Background ...... 45 3.2 Geological and Environmental Setting ...... 48 3.2.1 Geological Setting ...... 48 3.2.1 Environmental Setting ...... 49 3.3 Material and Methods ...... 51 3.3.1 Pre-Processing of Tide Gauge Data ...... 51 3.3.2 Pre-Processing of Vintage Aerial Photographs ...... 54 3.3.3 Pre-Processing of Satellite Images ...... 54

3.3.4 Change Detection from Time-Separated Remote Sensing Data in the Andros Tidal Flats ...... 55 3.3.5 Quantifying Change in the Configuration of the Andros Tidal-Channel Network Through Time ...... 58 3.4 Results ...... 59 3.4.1 Migration of the Boundaries of the Andros Intertidal Channelized Zone Through Time ...... 59 3.4.2 Changes Through Time in the Architecture of the Andros Tidal Channels at Triple Goose Creek ...... 62 3.5 Discussion ...... 64 3.5.1 Scytonema-Dominated Facies Respond Swiftly to Changing Sea Level ...... 64 3.5.2 Transgression in the Rock Record ...... 68 Chapter 4: Conclusion...... 74 References ...... 78

LIST OF FIGURES

Figure 1. (A) The Chagos Archipelago is situated in the central Indian Ocean and consists of eight isolated carbonate platforms ornamented by more than 55 individual atoll islands. This study considers two of these platforms. In the north, Peros Banhos (B), hosts 32 islands atop its reefal rim. The second site, Diego Garcia (C), is the southernmost atoll in the archipelago and ornamented by a single island which occupies the entirety of the atoll rim, except for its northern margin. The red box in (C) denotes the location of the U.S. military port and airfield complex which has been constructed on Diego Garcia in the last 50 years. Although human modification of the island has been concentrated in this sector of the atoll, infrastructure has been built elsewhere along the eastern limb of the atoll. The high fidelity of the vintage aerial photographs and modern satellite data is emphasized using representative imagery from Moresby Island, Peros Banhos, and South Bay, Diego Garcia (D-E & F-G, respectively)...... 6

Figure 2. (A) Emphasizes two possible definitions of an atoll-island coastline using a stretch of the lagoon-facing coastline of Moresby Island (Peros Banhos). The white-beach line (white line in A) denotes the interface between the beach and the high-tide mark. We do not gauge coastline shifts against this line as the swash zone is both highly dynamic and challenging to identify in time- separated remote sensing images because of tidal differences. Instead, this study assesses coastline change against the ‘brown soil’ line (yellow in A) which denotes the transition from beach sand to vegetated brown soil. (B) Shows the collection of GPS-constrained Google Street View imagery using a backpack-mounted sensor package. Shown here is Capt. Jon Slayer walking the coastline of Moresby Island. These 360° photogrammetric data were used to validate the coastline as digitized from WorldView-2 imagery (text for details)...... 13

Figure 3. Workflow for coastline extraction as applied to Ile Diamant (Peros Banhos). The “brown soil” line (text for details) is manually digitized from modern panchromatic WorldView-2 satellite imagery (A) and georectified vintage aerial photography (B). Binary representations of these coastlines are processed to emphasize areas of the island which have expanded versus retreated over the period of observation (C). Areas marked in yellow in this example have expanded during the period of observation. Purple demarks retreat. A geodesic-distance transform is applied to enumerate the distance of coastline migration in meters (D)...... 15

Figure 4. (A-B) As developed for Grande Mapou, an island situated on the northwestern rim of Peros Banhos Atoll, for the purpose of GIS analysis, the coastline of each island was partitioned into ocean-, lagoon-, and hoa-facing (yellow, pink, and cyan lines, respectively). Diego Garcia (C- D) hosts a single large island and therefore the hoa-facing category is not applied to this atoll. .. 17

Figure 6. Coastline-shift maps highlighting areas of land gained and lost between 1979 and 2015 for the islands of Peros Banhos. Hot colors denote areas of coastline retreat and cool colors denote expansion. Grayscale WorldView-2 satellite image acquired in 2015 is used as base layer. (A), (B), (C) and (D) illustrate that the ocean-facing coastline of several islands distributed on the western and northern rim show pronounced expansion, which has been delivered by the beachward migration of the brown-soil line. (A) shows the expansion of entire ocean-facing coastline, (B) and (C) are taken from northern rim, where pronouncing expansion could be mainly found on the ‘U’- shaped part of the coastline. (C) further shows a second mechanism for the seaward expansion of an island-embayment infilling. (D) emphasizes the overarching pattern of coastline migration in Peros Banhos whereby the ocean-facing coastline expands while the lagoon-facing coastline retreats. Meanwhile, the hoa-facing coastlines are particularly dynamic. The tips of the islands in (E), (F) both retreated by more than 30 m. (G) highlights the rapid accretion of sand spits. Both (A) and (H) exhibit marked erosion of land tips...... 29

Figure 7. Coastline-shift maps highlighting areas of land gain (blue) and loss (red) between 1963 and 2013 in Diego Garcia. High-resolution satellite imagery acquired in 2013 is used as the base layer. (A) through (C) illustrates how tranquil embayments on the lagoon-facing coastline of the island are highly dynamic and swiftly infill through time, thereby increasing the area of the island. Cases of land lost on Diego Garcia tend to be on ocean-facing coastlines (D, E, and F). In rare cases, however, ocean-facing coastlines do prograde seaward, such as resolved in (H) and (I), accompanied by a loss of land on the lagoon side. Seaward expansion of Simpson Point (H) is particularly pronounced...... 30

Figure 8. Trends in coastline migration versus island shape and size for Peros Banhos. (A) Plots the 10 islands that have expanded in area between 1979 and 2015, whereas (B) considers those which have contracted. Note consistency in behavior for both cases – circular islands are less dynamic than those which have an elongate shape. The propensity of an islands’ coastline to migrate is unrelated to its size...... 31

Figure 9. Differential behavior of island coastlines for Peros Banhos. (A) Satellite image of Peros Banhos Atoll. Each atoll islands for which paired remote sensing data exist is given a unique integer identifying code. For instance, Ile Diamant in the northwest of the atoll is assigned the number ‘1’, Grande Mapou ‘2’, and so on. These integers provide a key for the islands in (B) through (E) which have been abstracted to squares for ease of interpretation. Per the key in (B), small squares represent islands which are <20 ha in area, whereas the larger squares represent those >20 ha. The color of the squares encodes the degree of expansion (blue) or retreat (red) of each island over the period of observation. (B) Describes the behavior of the entirety of the island coastlines, whereas (C) considers only ocean-facing coastlines, (D) lagoon-facing, and (E) hoa-facing...... 33

vii

1979-2015 versus reef-flat width. The wave energy incident on each coastline is denoted in color and circles denote ocean-facing coastlines and squares lagoon-facing. More energetic coastlines are fronted by wide reef flats and retreat through time...... 36

Figure 11. The hydrodynamic setting of Diego Garcia. (A) Map of modelled wave energy incident on the ocean-facing coastlines of the island. (B) Plots of coastline migration for the period 1963- 2013 (y-axis) versus reef-flat width (x-axis). The wave energy incident on each coastline is denoted in color and circles denote ocean-facing coastlines. Unlike the case for Peros Banhos, there is no clear relationship between the hydrodynamic energy incident on the coastline of Diego Garcia, the width of the reef flat, and the change in island size through time...... 36

Figure 12. Human impact on coastline erosion. Three categories of coastline were considered between the two atolls. (A) Stretches far from human modification on both Diego Garcia (orange) and Peros Banhos (gray), and stretches of coastline in the immediate vicinity of human modification on Diego Garcia (blue). (B) Shows the average coastline migration per decade of the three categories. Error bars denote one standard deviation around the mean. The behavior of the coastline in the vicinity of human activity on Diego Garcia is significantly different (p = 0.05) to that of the undisturbed stretches on the same atoll and those considered from Diego Garcia – text for details...... 40

Figure 13. A compilation of data describing the behavior of atoll islands in the Indian and Pacific Oceans (y-axis) versus size of those islands (x-axis). Any decadal change in area less than 3% of the total island (blue zone bounded by dashed-horizontal blue lines) is generally considered as not representing a meaningful change – text for details. The vertical dashed black line separates the 559 islands <20 ha in area versus the 171 islands larger than this threshold. The pie charts specify that 71% of islands <20 ha do not meet the 3% criterion for meaningful change, which increases to 88% for those islands >20 ha. Regardless of ocean basin and biogeophysical setting, small islands are considerably more dynamic than large...... 42

Figure 14. (A) Map of rates of sea level rise for the tropical and subtropical oceans compiled by the data from satellite altimetry for the years 1993 through 2018 (AVISO). Hot colors denote rapid rates of rise and cool colors emphasize slower rates. Plotted in (B) is the rate of sea level rise for the waters offshore the considered atoll islands versus their decadal change in area. Note lack of correlation between the two variables, suggesting local factors to be a more important determinant of island dynamics in the considered time period than change in eustatic sea level...... 44

viii

platform. (B) Shows the wind rose capturing wind speed and direction extracted from QuikSCAT satellite observations for the period January 2013 to January 2014. (C) Three major hurricanes passed in close proximity to Andros between 1985 and 2018: Hurricane Andrew (Category 5) in 1992, Michelle (Category 4) in 2001, and Matthew (Category 5) in 2016, and are demarked in pink, yellow and green, respectively...... 51

Figure 18. Analysis of the migration of the boundaries of the intertidal-channelized zone was conducted in four focus areas distributed from North to South along the western coastline of Andros Island. The focus areas capture variation within the channelized zone and as well as varying orientation of the coastline with respect to prevailing hydrodynamic energy...... 57

Figure 19. An example (A-B) of applying the Digital Shoreline Analysis Systems (DSAS) for measuring the migration of boundaries. Baseline was constructed parallel to the shoreline with a 30m interval (A). The distance of migration was calculated automatically according to the position of the transects. Different colors for the transects denote net shoreline movement in units of meters (B)...... 57

Figure 20. Black boxes in (A) shows the locations of the four studied areas in the Triple Goose Creek. (B-E) Channel bifurcation hierarchy in the Triple Goose Creek. Channels were classified according to their number of bifurcations that link back to the ‘main’ Order 1 channel which feeds the network from the coastline (text for details) ...... 59

(E). The timing of hurricanes, over-plotted with horizontal blue lines, appear to have little impact on patch patterning, whereas increases in sea-level, first in 2001 and again in 2011, appear to exert meaningful control, leading to fragmentation of the Scytonema landscape (text for details)...... 62

Figure 23. Changes of the morphology of the channel network assessed in four focus areas within the Triple Goose Creek sector of the Andros Island tidal flats in the period 1943 through 2018. (A) and (B) map the channel bifurcations, color-coded by their hierarchy (text for details) for the four areas in 1943 and 2018, respectively. (C) Charts the cumulative length of the channels gained and lost in this time period, by bifurcation hierarchy. Note that the greatest changes are observed in the two central focus areas where the tidal flat is broadest (D) and where the coastline is east facing (E)...... 64

Figure 25. (A) Shows the location of the Andros Town where the rainfall data were collected. (B) Plots the annualized rainfall amount from 2009-2018...... 68

Figure 26. (A-B) shows the locations of the three hypothetical cores in both vertical and horizontal respects in the intertidal-channelized zone: Core 1 locates in the proximal location of the coastline; Core 2 in the middle point of the intertidal-channelized zone; Core 3 locates in the distal portion of the intertidal-channelized zone. Based on the results of our time-separated remote sensing, (C) shows the idealized rock record in the three-hypothetical cores prior to the initiation of the earliest transgression, during the transgression, and immediately after it...... 72

Figure 27. A core sample collected in one of the abandoned channels in the Triple Goose Creek (A). In 1943(B), the channel was active but abandoned in 2018 (C). (D-E) shows the 120 cm-long sediment core with photograph and description, respectively. In the core, laminated mudstone (F) extends from 0-80 cm and overlies gastropod-rich wackstone (G). These gastropod shells (in G) are considered as remnants of the channel lag...... 73

LIST OF TABLES

Table 1. Summary of the behavior of the 21 considered islands in the Chagos Archipelago...... 25

Table 2.Annualized rates of coastline change for Peros Banhos for the period 1979-2015 and for Diego Garcia (1963-2013) as determined from time-separated remote sensing...... 26

Chapter 1: Introduction

The concept of “comparative sedimentology” was established by Charles Lyell and

Johannes Walther in the 19th Century and facilitates a comparison of facies patterns within and between geological periods. This discipline allows geologists to reduce uncertainty and increase accuracy in their interpretation of the subsurface (Ginsburg, 1974; Purkis and

Harris, 2016). The expansion of such comparative research on facies in carbonates improved the understanding of coastal-zone dynamics in the early 21th century, and has the potential to yield insight into geological processes recorded in the rock record. Furthermore, with about 40% of the world’s human population living within the coastal zone, the response of coastlines to rising seas is a topic of pressing interest. This study considers two coastal settings in pure carbonate depositional systems, but with radically contrasting energy regimes. First, the atoll islands of the Chagos Archipelago, which are supratidal grainy deposits, rimmed by reefs, and, because they are situated on the margins of isolated carbonate platforms in the Indian Ocean, are subject to high-energy conditions. The second case study is the low-energy inter- and supratidal muddy deposits of the tidal flats which mantle the west coast of Andros Island. Protected by the expansive platform interior of the

Great Bahama Bank to the west and Andros Island immediately to the east, these deposits accumulate in a comparatively quiescent environment. In addition, they experience episodic disturbance by hurricanes. By contrasting the two case studies, the project aims to examine the dominant controls on inter- and supratidal facies architecture for a platform-margin and platform-interior setting, with a particular emphasis on the response of the two systems to contemporary sea level rise.

Chapter 2: High-Energy Facies Dynamics in the Central Indian

Ocean

2.1 Background

Coastlines are dynamic environments. Their morphology is influenced by a host of controls, including sea level, on timescales of minutes to millennia. Not least because of our desire to live by the sea, the dynamics and drivers of coastal change have been studied intensively (Richmond, 1993; Kaluwin and Smith, 1997; Dickinson, 1999; Masselink and

Pattiaratchi, 2001; Woodroffe, 2008; Collen et al., 2009; Webb and Kench, 2010; Rankey,

2011; Purkis and Klemas, 2011; Taylor and Purkis, 2012; Andréfouët et al., 2013; Purkis et al., 2016; Duvat and Pillet, 2017; Duvat, 2020; Newnham et al., 2020). Notwithstanding raised atolls, those that have been tectonically uplifted, atoll islands are low and flat with maximum elevations typically in the range of only a few meters. Being only slightly above sea level carries several implications. First, atoll islands are vulnerable to extreme events such as catastrophic storms, as well as to global environmental change such as sea-level rise, in particular. On the latter point, though, it has been suggested that some (but not all) coral islands might be morphologically resilient to rising sea level (Kench et al., 2005;

Woodroffe, 2005, 2008; Webb and Kench, 2010; Kench et al., 2015; McLean and Kench,

2015; Beetham et al., 2017; Tuck et al., 2019), though some authors have argued to the contrary (Hubbard et al., 2014; Storlazzi et al., 2015). In addition, atoll islands have little land area and limited natural terrestrial resources, e.g., climate-sensitive supplies of fresh groundwater (Bailey, 2015; Barkey and Bailey, 2017; Falkland and White, 2020). Finally, terrestrial plants build the base of the entire inland atoll ecosystem. The decomposition of

the terrestrial plant material (leaves, stems, rots, etc.) forms a layer on the soil surface

(Boberg, 2009; Coleman et al., 2017) herein referred to as ‘brown soil’. Because of its biological origin, expansion of brown soil areas will occur on timescales determined by the lifecycle of the constituent plant species. As more than 99.8% of the terrestrial plant species have a finite tolerance to salt (Flowers and Colmer, 2015), it is anticipated that repeated exposure to seawater will lead to rapid ecosystem collapse and thus ‘brown soil’ contraction will reflect rates of plant death and soil structure degradation.

Whereas atoll islands are rare in the tropical Atlantic, they are common in the Pacific and Indian Oceans, where they constitute entire countries or vast overseas territories of continental states, serve as home to hundreds of thousands of inhabitants, and house critical and strategic infrastructure. Out of 709 atoll islands which have so far been analyzed to quantify shoreline dynamics, 533 islands are located in the Pacific and only 176 in the

Indian Ocean (Duvat, 2019). Of the work accomplished in the Indian Ocean, the Maldives

Archipelago has received the most attention (Aslam and Kench, 2017). Even from this portfolio of studies, only one (at the scale of entire atolls), Nadikdik Atoll in the Pacific, is an uninhabited site, thereby allowing for the dynamics of an anthropogenically-unaltered atoll to be observed (Ford and Kench, 2014). Although it has long been recognized that artificial modification of the coastal zone can strongly influence shoreline migration (Ford,

2012; Duvat and Pillet, 2017), this latter study emphasizes that even atoll islands which are uninhabited by humans can also display meaningful natural dynamics.

The paucity of studies which consider both uninhabited and inhabited islands in the same archipelago and under the same biophysical controls frustrates efforts to disentangle the human effects of coastline erosion triggered by inappropriate intervention, such as the

construction of seawalls that can starve beaches of necessary sediment supply, from the natural dynamics that should be anticipated for atoll islands exposed to normal marine conditions. A full and thorough understanding of the natural dynamics of these systems in a period of relatively static sea level, as compared to the rapid rise witnessed in the Early to Mid-Holocene, can provide a critical baseline as to the lower-end of island reconfiguration that should be anticipated in the present era of accelerated sea level rise.

In this study, we have assembled a suite of archive aerial photographs (negatives) for the Chagos Archipelago (Indian Ocean) (Fig. 1A). These photographs were acquired in the

1960’s and 1970’s and have an effective resolution finer than 0.2 × 0.2 m. In this study, we consider two atolls. First, Peros Banhos (Fig. 1B), which represents one of the very few atolls globally where natural island dynamics can be quantified at the scale of a whole atoll

– withstanding just one of the 32 islands, the Peros Banhos has never been settled by humans. This atoll represents an excellent natural laboratory because its islands are near- evenly distributed around the atoll rim, thereby allowing for nuanced differences between windward and leeward positions to be examined. The second atoll that we consider is Diego

Garcia (Fig. 1C). While the majority of the coastline of Diego Garcia is unmodified, a U.S. military port and airfield complex has been constructed in the last 50 years on the western limb of the atoll which now occupies approximately 10% of its circumference (see red box in Fig. 1C). The coastline dynamics of Diego Garcia have been considered by both

Hamylton and East (2012) and Purkis et al. (2016), and therefore the trajectory of the atoll with regard to prevailing natural and anthropogenic forcings is well understood.

Capitalizing on this previous work for Diego Garcia and expanding it to include the uninhabited islands of the adjacent Peros Banhos Atoll allows the dynamics of atoll islands

in their natural state (Peros Banhos) to be contrasted with dynamics of an anthropogenically-disturbed system (Diego Garcia), while avoiding bias introduced by varying ocean climate since both sites are closely situated in the same archipelago. The comparison is further strengthened by the subequatorial position of Chagos which prevents its atolls from being directly impacted by cyclones – events that can cause radical local divergence of coastline dynamics, even for closely situated islands (Kench and Brander,

2006a; Ford and Kench, 2016; Duvat et al., 2017). Note, however, that the Chagos atolls, like those of the adjacent Maldives (Aslam and Kench, 2017), are exposed to distant-source swells originating in the southern Indian Ocean and therefore not devoid of cyclonic influence.

The goals of the study are twofold: First, using remote sensing data spanning more than 30 years, to quantify whether the islands of Peros Banhos (Fig. 1D), an atoll which has not been disturbed by humans, behaves differently from Diego Garcia (Fig. 1E).

Second, using a hydrodynamic model for both atolls, to explore the physical controls driving coastline change of the two systems. Answers to such questions are urgently needed

– there is reasonable doubt as to whether many atoll islands globally will remain inhabitable in the face of rising sea level and decreasing sediment supply resulting from the wholesale collapse of coral reef systems wrought by ever more frequent marine heatwaves (Perry and Morgan, 2017; Perry et al., 2018).

2.2 Regional Setting

The Chagos Archipelago (Fig. 1) is part of the British Indian Ocean Territory (BIOT) and lies in the center of the Indian Ocean, occupying a southern extension of the

Laccadive–Maldives Ridge. There are 55 atoll islands distributed atop the eight Chagos atolls. These islands are fewer and smaller than those of the Maldives or Laccadive

Archipelagos. Like coral atoll islands globally, the Chagos Islands are low lying (only 2-5 m above sea level) and are typical coral cays constructed of limestone with underlying freshwater lenses sustained by high rainfall (Sheppard et al., 2012). Of the 21 Chagos islands considered by this study, only the one atop Diego Garcia Atoll is inhabited.

Whereas Diego Garcia, which is situated in the southeast of Chagos, is one of the smallest atolls of the archipelago (28 sq. km), it hosts the largest emergent landmass which accounts for 60% of its total island area. By contrast, Peros Banhos Atoll covers 463 sq. km, of which only 9 sq. km is accounted for by its islands, all of which are situated on the atoll rim. The Chagos climate is tropical and moderated by the trade winds (Eisenhauer et al.,

1999). Until very recently, the archipelago’s coral reefs were noteworthy for their health and vitality which are aided by the declaration of the Chagos Marine Protected Area in

2010, which, at the time, was the world’s largest (Purkis et al., 2008; Sheppard et al., 2012).

This said, the reefs of the Chagos have suffered gravely in the last five years by virtue of pervasive coral bleaching (Sheppard et al., 2017; 2020).

There is only one ‘research-grade’ tide gauge maintained in the Chagos Archipelago and this is situated in Diego Garcia. Monthly-mean data compiled by Purkis et al. (2016) resolve the rate of relative sea-level rise (i.e. that rate experienced by the atoll) as 5.44 mm yr−1 for the period 1988–2000 and 5.96 mm yr−1 for the period 2003–2014. These relative

rates of rise are higher than the global eustatic signal as resolved by Dunne et al. (2012) for the central Indian Ocean of 3.12 mm yr−1 for the period 1988-2011. The difference here can likely be attributed to the subsidence of the Chagos Archipelago (i.e. there is a difference of approx. 2 mm yr-1 between the rate of change of relative verses eustatic sea level). Tides in the Chagos are semi-diurnal with a maximum magnitude of 1.5 m. As captured by the meteorological station at the Diego Garcia airfield, the Chagos Archipelago is situated in the trade-wind belt from May to September during which time the wind blows strongly from the southeast. Low velocity cyclonic winds characterize the rest of the year.

Whereas these climate patterns exert some control of the local wave fields, the archipelago, like the neighboring Maldives, also receives distant-source southerly swells originating in the southern Indian Ocean.

The shallow-subsurface of the Chagos atoll-islands has yet to be examined in detail

(likely because of the logistical and legal challenges of importing drilling equipment to the archipelago) and therefore the genesis of the islands is poorly constrained. The is not a complete deficit of data, however. A comprehensive study conducted by Eisenhauer et al.

(1999) reports the fossil reefs that underlie many of the Chagos islands to have mostly accumulated at a time coinciding with sea-level fall during the Mid-Holocene. During this time, the relaxation of the continental crusts in glaciated areas and the reorganization of the period after the waning of the major ice caps caused a redistribution of water mass globally

(e.g. sea level rises relative to Caribbean islands and fall around Indian Ocean and

Australian continental margin). Conversely, Kench et al. (2005) and Perry et al. (2013) provide evidence that the atoll islands of the Maldives, a good analogue to those of the

Chagos because of their proximity, accumulated earlier and with the sedimentary infilling

of lagoons, when sea level was rising rather than falling. On the basis of the available evidence, we can be confident that the Chagos islands are “pinned” by Pleistocene and younger aged reefal deposits. Further, it is evident that after the last interglacial, coral reoccupied the raised atoll margins of both the Chagos and Maldives Archipelagos and grew vertically, keeping up with sea-level rise to deliver the modern atolls (Droxler and

Jorry, 2020).

Almost all of the Chagos islands are covered with coconut groves (e.g. Barringtonia asiatica, alophyllum inopyllum, plus Cocos nucifera in Diego Garcia). Through the eighteenth and early nineteenth centuries, coconut oil became the primary produce of

Chagos, whose islands then gained their characteristic skyline of palms (Fig. 2).

Throughout the archipelago, plantations caused the native vegetation to be cleared and the coconut oil became so successful that the archipelago became known as the Oil Islands, prior to the industry’s collapse and abandonment of the coconut plantations. A few islands, notably on the Great Chagos Bank and Peros Banhos, remained too inaccessible for regular use, and escaped coconut planting. As a result, these, typically small islands continue to boast native hardwood trees and other original plants, together with high densities of bird species which were lost from the heavily-planted islands due to the introduction of rats

(Sheppard, 2016; Graham et al., 2018). Boosted by rat-eradication programs, seabird species, such as Onychoprion fuscatus, Anous tenuirostris and Sula sula, now inhabit 30 islands covering Chagos islands. Most of breeding seabirds are recorded in the northern and eastern Perhos Banhos islands including Ile Parasol, Ile Longue, Petite Bois Mangue,

Petite Coquillage, and Grande Coquillage (McGowan et al., 2008).

2.3 Material and Methods

2.3.1 Pre-Processing and Georectification of the Vintage Aerial Photographs

This study was facilitated by the discovery of a collection of vintage aerial photographs acquired over Diego Garcia in 1963 and over Peros Banhos in 1979. The former atoll was imaged in its entirety, whereas for Peros Banhos, a small handful of islands in the southwest of the atoll were not photographed (shaded gray in Fig. 1B) and therefore could not be included in this study. The negatives of the vintage aerials for both atolls were scanned at 3,000 pixels-per-cm to yield digital data with an effective spatial resolution of <0.2 m. Although the spatial resolution of the vintage photographs is adequate, they were acquired without any geo-information and thus suffered from geopositioning inaccuracies. Such inaccuracies can be corrected via georectification, a transformation process used to project an unpositioned (or poorly positioned) historical image onto a known coordinate system. This procedure involves pairing the archive data with a well- positioned satellite image and selecting points on the ground common to both. These locations become reference points in the subsequent warping of the unpositioned image onto a coordinate system. Although this study will compare the vintage photographs to modern satellite imagery, the latter was not used for georectification in order that it remained independent from the historical data. Due to differences in the data available to guide georectification of the vintage imagery, the procedure for Diego Garcia diverged from that used for Peros Banhos, as detailed in the next section.

As conducted by Purkis et al. (2016), the vintage photographs for Diego Garcia were georectified against a 2003 aerial survey conducted by the Naval Facilities Engineering

Command (NAVFAC). The spatial resolution of the NAVFAC mosaic is 0.15 m and,

through comparison with recent satellite imagery, its geo-positioning was deemed to be precise. To quantify change through time, the georectified aerials were paired with panchromatic WorldView-2 (then DigitalGlobe Inc., now Maxar Inc.) images which have a spatial resolution of 0.3 m. Suitable features on the island of Diego Garcia consisted of forest clearings, as well as freshwater ponds and intertidal channels. Farther offshore, points used for georectification were positioned on coral heads in the lagoon and on the reef flat. Since it is coastal shifts that we seek to quantify, features lying along the shorelines were purposely avoided as using these would risk that actual shifts in the shoreline would be explicitly compensated for by the warping of the imagery during georectification. Since NAVFAC imagery does not exist for Peros Banhos, georectification of the vintage photographs from this atoll was conducted against differential GPS (dGPS) measurements made throughout its islands during a 2015 field campaign. Via visual inspection of the paired vintage (1979) and modern (2015) imagery, points were selected in the interior of all the Peros Banhos islands which could be easily recognized in both datasets and were presumed to have been spatially invariant in the 36 years which had elapsed between the image acquisitions. These points were visited during fieldwork, the position of each logged using dGPS, and these data used to georectify the vintage photographs.

2.3.2 Pre-Processing of Modern WorldView-2 Satellite Imagery

The WorldView-2 data for Diego Garcia were acquired on November 8th, 2013 and those for Peros Banhos on March 3rd, 2015. These datasets were selected because of their low cloud cover, calm seas, and the fact that the times of acquisition both corresponded to low tide. To deliver the maximum possible spatial fidelity, the sub-meter resolution of

WorldView-2’s panchromatic band was merged with the coarser resolution multispectral bands. This process, termed ‘pansharpening’ was accomplished using ENVI software (v.

4.8) and delivered a four-band image (near IR, red, green and blue) with 0.5 m pixels.

WorldView-2 delivers data in eight spectral bands, of which five are water penetrating. In order of increasing wavelength, these are the coastal blue band (400–450 nm), blue (450–

510 nm), green (510–580 nm), yellow (585–625 nm), and red (630 - 690 nm).

2.3.3 Coastline Extraction and Development of Coastline-Shift Maps

Following the protocols of Purkis et al. (2016), the coastlines of the islands of both

Diego Garcia and Peros Banhos were manually extracted from the vintage and modern imagery. Here, a division is made in the analysis between changes in the wooded “brown soil” coastline of the islands and the shift of “white” beach sands (Fig. 2A). This is an important partition because beaches are naturally dynamic, even on seasonal timescales

(Kench and Brander, 2006a, 2006b; Ford, 2012; Holdaway and Ford, 2019). We posit that migration of brown soil areas, by contrast, represents a long-term meaningful change in island morphology (Webb and Kench, 2010). This study seeks to quantify the latter and coastline position is therefore defined as the transition from brown soil to beach sand, as also delineated by Hamylton and East (2012) and Purkis et al. (2016). The accuracy of the coastlines manually extracted from the WorldView-2 imagery was assessed against field data which remained independent from the extraction process. These data took the form of dGPS points which were recorded every 5 secs. by a fieldworker walking around each of the studied islands using a backpack-mounted GPS system. Care was taken to walk the

“brown soil” coastline and GPS recording was halted at any point that this could not be achieved. In addition to GPS data, the fieldworker carried a backpack-mounted 3-D

photogrammetric array which continuously acquired 360° imagery. This latter dataset was subsequently processed to produce an interactive ‘Google Street View’ rendering of the island coastlines (Fig. 2B) which could later be explored to allow for classification of coastal vegetation and morphology. Acquisition of these image data was facilitated via a partnership between XL Catlin Seaview Survey and the Chagos Conservation Trust.

Figure 2. (A) Emphasizes two possible definitions of an atoll-island coastline using a stretch of the lagoon-facing coastline of Moresby Island (Peros Banhos). The white-beach line (white line in A) denotes the interface between the beach and the high-tide mark. We do not gauge coastline shifts against this line as the swash zone is both highly dynamic and challenging to identify in time-separated remote sensing images because of tidal differences. Instead, this study assesses coastline change against the ‘brown soil’ line (yellow in A) which denotes the transition from beach sand to vegetated brown soil. (B) Shows the collection of GPS-constrained Google Street View imagery using a backpack- mounted sensor package. Shown here is Capt. Jon Slayer walking the coastline of Moresby Island. These 360° photogrammetric data were used to validate the coastline as digitized from WorldView-2 imagery (text for details). As developed in an example from Ile Diamant (Fig. 3), which is an island situated on the north-western rim of Peros Banhos, graphical GIS representations of coastline shift between the vintage aerials and modern imagery were created through analysis of the extracted coastline vectors (Fig. 3A-B). Here, a four-step process is followed. First, the

coastline vectors were transformed from ESRI shapefile format to 1-bit per pixel binary rasters with pixel values of ‘1’ denoting the spatial extent of the atoll island. The rasters were created for both the vintage and modern coastline positions and have a spatial resolution of 2 m. Second, matrix arithmetic was used to subtract the modern binary image from the vintage binary to yield a product capturing areas of island change. Land areas which have retreated during the period of observation were assigned pixel values of ‘-1’

(purple in Fig. 3C). Areas which have expanded, meanwhile, were assigned values of ‘1’

(yellow in Fig. 3C). Third, a geodesic-distance transform was computed on this product raster which served to assign a number to each pixel that is the constrained distance between that pixel and the position of the vintage coastline. Fourth, multiplying these distances by the spatial resolution of the binary images (2 m) yielded a raster consisting of pixel values that spatially capture the distance of coastline migration in units of meters.

This raster can be visualized within GIS using a color ramp such that cold colors denote areas of coastal expansion and hot colors signify retreat (Fig. 3D). This workflow was followed for all the islands ornamenting Peros Banhos and the one large island of Diego

Garcia, and quantitative information pertaining to the migration of the island facies harvested.

In order to resolve the variability in the behavior of different precincts of the considered islands, the coastline of each was partitions into one of three categories (Fig. 4).

The first category is the ‘ocean-facing’ coastline, which faces the open ocean. Similarly, the ‘lagoon-facing’ category was applied to coastlines facing into the lagoon of the atolls.

The third category is ‘hoa-facing’ coastlines which are those inter-island channel-facing shorelines (i.e. neither ocean- nor lagoon facing, as developed in Fig. 4B – cyan text). For

the case of Peros Banhos (Fig. 4A-B), the islands are typically elongate with their principal axes aligned to that of the atoll rim. For this reason, the ocean- and lagoon-facing precincts of the islands tend to be substantially longer than the hoa facing. To eliminate the bias potentially caused by varying lengths of different precincts, relative changes in island size are presented in lieu of absolute changes. In contrast to Peros Banhos, Diego Garcia hosts a single large island and therefore the hoa-facing category is not applied for this atoll (Fig.

4C-D). The other difference in our treatment of Diego Garcia is that the area of the atoll modified to accommodate the military facility is excluded from analysis of coastline change. The radical dredge-and-fill modification of this area of Diego Garcia has been detailed by both Hamylton and East (2012) and Purkis et al. (2016) and is beyond the scope of this study.

Figure 4. (A-B) As developed for Grande Mapou, an island situated on the northwestern rim of Peros Banhos Atoll, for the purpose of GIS analysis, the coastline of each island was partitioned into ocean-, lagoon-, and hoa-facing (yellow, pink, and cyan lines, respectively). Diego Garcia (C-D) hosts a single large island and therefore the hoa-facing category is not applied to this atoll. 2.3.5 Sources of Error from Coastline Extraction

Our evaluation of positioning error follows the protocols outlined by Purkis et al.

(2016). In the following two paragraphs, two major systematic errors and solutions to minimize them are introduced in detail.

One source of error arises due to mis-rectification errors. Although maximum effort was made to ensure precise georectification of the vintage aerial photographs, precision cannot be wholly achieved. Appraising rectification errors is important since they propagate into measurements of coastline dynamics. Compared to the archive aerials, a much higher level of confidence can be held in the positional accuracy of the WorldView-

2 imagery as the orbit of this modern sensor is precisely controlled and the geo-location of the imagery routinely validated by Digital Globe Inc., the satellite's operator. Accurate positioning of the WorldView-2 imagery can then be verified through cross-comparison of

50 stable points (road junctions, buildings, etc.) selected from the 2003 NAVFAC aerials, which yielded a root mean square error of less than 1m.

Another source of error comes from coastline extraction in areas where trees extend over the boundary between brown-soil line and white sand beach. To mitigate this source of error, these areas were manually digitized with reference to the dGPS tracks acquired in the field. Purkis et al. (2016) quantitatively assessed the uncertainty of extracting the brown soil vector in Diego Garcia by comparing the satellite-derived coastline with field GPS tracks (See their Fig. 6). This assessment indicated the average offset of the former coastline from its true position in the landward direction to be 1.8 m and 2.4 m in the seaward direction. These results suggest that the techniques employed to derive the position of the brown soil coastline from the remote sensing images has a maximum uncertainty on the order of 3 m, equal to the breadth of 2 pixels in the data layers used for the analysis.

Coastline change within 3 m (±3 m) are considered to be ‘stable’, whereas changes ≤ or ≥3 m are defined as ‘dynamic’. Such error range is in accordance with the results shown in other equivalent studies (e.g. Ford, 2012; Duvat et al. 2017; Duvat and Pillet, 2017; Kench

et al., 2018). Repeating the analysis for Peros Banhos confirmed the uncertainty to be the same for the islands of Peros Banhos.

2.3.6 Hydrodynamic Modelling of Peros Banhos and Diego Garcia

Given that hydrodynamic energy exerts substantial control on island architecture

(Ekebom et al., 2003; Hernández-Cruz et al., 2006; Taylor and Purkis, 2012), we simulated the prevailing long-term wave exposure regime (see 2.3.6 for detailed wave information) for the Chagos Archipelago. This computer simulation was accomplished using the empirical routines proposed by Chollet and Mumby (2012), and subsequently adopted by

Purkis et al. (2015) and Perry et al. (2015), which are conditioned using daily wind speed and direction retrieved from QuikSCAT. These satellite data were validated against the meteorological station at the Diego Garcia airfield and deemed accurate. Fetch lengths – i.e. the horizontal distance over which wave-generating winds blew – were computed for the entire archipelago with reference to the Millennium Coral Reef-Mapping Project

(Andréfouët et al., 2006). Here, reefs shallower than 3 m were considered to interrupt wave propagation, with the result that the majority of the Great Chagos Bank situated in the center of our model domain (and the largest atoll on Earth), did not form a hydrodynamic barrier to Peros Banhos, since its rim languishes considerably below fair-weather wave base. Wave exposure (J/m3) was calculated using the same method described by Ekebom et al. (2003) and by Chollet and Mumby (2012).

2.3.6.1 Computing Fetch Length

The coastlines of the islands of Diego Garcia and Peros Banhos were divided into series of nodes, each separated by 100 m. Using code developed in Matlab (MathWorks v.

2019a), each node was visited, and the fetch length computed in 36 compass directions

following Purkis et al. (2015). In this step, fetch is calculated with reference to the

Andrefouët et al. (2006) reef map for the archipelago that captures the position of emergent features, as well as submerged areas reaching within 3 m of the sea surface.

2.3.6.2 Calculating Hydrodynamic Exposure

For each coastline node, wind speed and direction were derived from QuikSCAT satellite scatterometer data for the period 1988 to 1998. With fetch and a climatology of wind speed and wind direction now parameterized for each node, spectral wave height and period could be computed following Chollett and Mumby (2012). Described by Ekebom et al. (2003), the wave prediction method is illustrated as follows. For a given wind speed and unlimited wind fetch, there is a fixed limit to which the average wave height and period will grow. At this limiting condition, the rate of energy input from the wind to the waves is balanced by the rate of wind-energy dissipation due to wave breaking and turbulence.

This condition, known as a fully-developed sea, is used for the development of many standardized wind-wave spectra (Sorensen, 2005; Ekebom et al., 2003). Fully-developed sea conditions are no longer fetch-limited. We used, for fetch-limited conditions, a threshold in which the non-dimensional fetch ξ was shorter than 38,590 m.

2 ξ=g * F/U10 <38,590 (1)

2 where g is the acceleration due to gravity (9.81 m/s ), F is the fetch (m), U10 is the wind speed at an elevation of 10 m in ms-1.

If the calculated spectral wave height Hm0 (Eq. 2) and peak spectral period Tm (Eq. 3) reached fully-developed sea conditions, Eqs. (4) and (5) were used instead (Kahma and

Calkoen, 1992). The error in predictions caused by changes in wind speed and direction

over large fetch distances must thus be accepted, but are considered negligible in the context of this study which seeks information over long time and space scales.

1.1 0.45 Hm0=0.00082 * U10 * F (2)

where Hm0 is the spectrally based significant wave height (m), U10 the wind speed elevation at 10 m (m/s). The period of the spectral peak is calculated by the equation:

0.46 0.27 Tm=0.087 * U10 * F (3)

If wind and fetch converge to deliver fully-developed sea conditions, the maximum values of wave heights and periods are calculated following Kahma and Calkoen (1992):

2 Hm0=0.034 * U10 (4)

Tm=0.810744 * U10 (5)

The total energy of a wave system is the sum of its kinetic and potential energy, where the kinetic energy is the part due to water particle velocities associated with water motion, and potential energy results from the part of the fluid mass above the wave crest.

Since wave energy (WE) is proportional to wave height:

2 WE=1/16ρgHm0 (6)

3 where ρ is the sea water density (1,030 kg/m ) and Hm0 is the wave height (m). When all equations have been performed for each of the 3,652 days (10 years) simulated, then the average energy, E, in joules can be calculated as the geometric mean of the individual energy estimates for each node of the coastlines of Diego Garcia and Peros Banhos.

2.4 Results

2.4.1 Changes Through Time in the Areas of the Peros Banhos and Diego Garcia Atoll

Islands

For each of the 20 Peros Banhos islands for which paired remote sensing data exist, and for the single island of Diego Garcia, per-island statistics were extracted via GIS analysis. We assembled data to describe the per-island net changes in island area for the periods of observation (1979-2015 for Peros Banhos; 1963-2013 for Diego Garcia). For each island, net areal change was also computed as a percentage of the overall size of each island, and as a rate of change, taken to be the change in island area per decade (Table 1).

These data reveal that whereas some islands have expanded in area through time and others contracted, the decadal percentage changes are exceedingly small, at <1% when averaged across all of the considered islands, except for the outlier of Ile Fouquet (Peros Banhos), which contracted in area by -6.55% per decade of observation (Fig. 5A-D). This island, which is situated on the southern rim of Peros Banhos, lost 19.1% of its area from the ocean-facing coastline and 13.6% from its hoa-facing coastline between 1979 and 2015, while its neighboring islands behaved in a manner consistent with the broader dataset. The net areal reduction in island area over the period of observation for the 20 uninhabited

Peros Banhos islands is 0.004%, increasing to 0.900% for Diego Garcia. However, it is important to realize that these low values of net areal change computed over entire islands do not exclude the possibility of substantial coastline dynamics, such as extension in one part of an island, offset by erosion in another (e.g. Aslam and Kench, 2017). Such local nuances are captured by our separate analysis of ocean-, lagoon-, and hoa-facing coastlines.

A number of recent studies have coalesced around the consensus that any change in island

area <3% of the total land area lies within the error of time-separated remote sensing

(McLean and Kench, 2015; Duvat & Pillet, 2017; Duvat et al., 2017; Duvat, 2019). Placing our Chagos data in this context emphasizes the net stability of the atoll islands during the period of observation. The average areal change, however, is not representative of the high degree of coastline dynamics (Fig. 6 & 7) which can be recognized when the island coastlines are partitioned according to whether they are lagoon-, ocean-, of hoa-facing (Fig.

4).

Figure 5. Ile Fouquet, which is situated on the southern rim of Peros Banhos (A-B), lost nearly one quarter of its ocean- and hoa-facing coastlines between 1979 and 2015. The green line in (B) depicts the position of the coastline in 1979, whereas the orange line depicts its position in 2015. (C) and (D) emphasize the extraordinary erosion that has occurred in the span of 36 years. The ocean-facing coastline has lost 4,620 sq. m in this time period, equivalent to 58% of the area of the island. The hoa-facing coastline, (D), lost 3,293 sq. m. Annualized average rates of coastline migration are assembled in Table 2. Of the coastlines which have expanded during the period of observation, it is those which are hoa- facing that show the highest rates of migration for Peros Banhos (average migration rate over 20 islands of 0.24 m/yr., versus 0.16 m/yr. for lagoon- and 0.18 m/yr. for ocean-facing coastlines). Of the coastlines which have retreated during the period of observation, it is those which are ocean-facing that show the highest rates of movement for Peros Banhos

(average migration rate over 20 islands of -0.34 m/yr., versus -0.15 m/yr. for lagoon- and

-0.28 m/yr. for hoa-facing coastlines). As emphasized in Table 2, and in contrast to Peros

Banhos, on Diego Garcia the lagoon-facing coastline has expanded more rapidly than the ocean-facing (0.27 vs. 0.21 m/yr., respectively). Rates of retreat are also more rapid on the lagoon-facing shoreline of Diego Garcia than on the island’s ocean-facing shoreline at 0.37 m/yr. and 0.25 m/yr., respectively, data which, in aggregate, suggest the lagoon-facing coast to be considerably more dynamic than the ocean-facing side. This said, and as emphasized by Purkis et al. (2016), the most dynamic precinct of Diego Garcia is the lagoonal embayments in the south of the atoll (emphasized in Fig. 7B). However, the dynamic behavior of these embayments serve to mask lethargic coastline changes elsewhere on Diego Garcia. For instance, recomputing annualized changes for this atoll island with its southern embayments omitted, reveals the lagoon-facing coastline to be retreating more slowly than the ocean-facing coastline (-0.02 m/yr. vs. -0.06 m/yr., respectively). When the degree of expansion and retreat is averaged across all the islands of Peros Banhos, and across Diego Garcia, both atolls emphasize that the total land area is almost static over the period of observation because area lost is nearly balanced by area gained. The net annualized rate of change is -0.02 m/yr. for Peros Banhos and −0.07 m/yr. for Diego Garcia.

Table 1. Summary of the behavior of the 21 considered islands in the Chagos Archipelago.

Decadal Time Average Atoll Island Net Areal Change Areal Period Areal Island Area (ha) (%) Change Change (ha) (%) % Petite Mapou 1.10 -0.03 -3.07 -0.85 Ile Finon 1.27 -0.03 -2.09 -0.58 Ile Fouquet 1.85 -0.57 -23.58 -6.55 Ile Manon 2.20 0.04 1.74 0.48 Unnamed 2.40 -0.07 -2.67 -0.74 Island Ile Verte 3.75 0.04 0.97 0.27 Mapou du 6.74 0.14 2.15 0.60 Coin Ile Parasol 7.74 0.11 1.50 0.42 Petite Bois 8.01 0.06 0.08 0.22 Peros Mangue 1979- Banhos Grande Bois 2015 11.83 0.23 1.99 0.55 -0.004 Mangue Ile Longue 17.81 0.28 1.62 0.45 Grande 18.74 -0.06 -0.34 -0.09 Mapou Ile Passe 20.69 -0.21 -1.00 -0.28 Ile Manoel 29.40 0.15 0.51 0.14 Moresby 30.87 1.31 4.44 1.23 Island Petite Soeur 46.22 -0.85 -1.81 -0.50 Grande Soeur 55.24 -0.19 -0.35 -0.10 Ile Diamant 83.26 1.72 2.11 0.59 Ile Poule 89.31 -1.18 -1.31 -0.36 Ile Pierre 118.12 -0.92 -0.77 -0.21 Diego 1963- ̶ 2124 -19.49 -0.90 -0.18 -0.900 Garcia 2013

Table 2. Annualized rates of coastline change for Peros Banhos for the period 1979-2015 and for Diego Garcia (1963-2013) as determined from time-separated remote sensing. For Diego Garcia, 1entire coastline including southern embayments and 2excluding embayments.

Entire Ocean- Hoa-facing Migration Lagoon-facing Coast Atoll facing Coast Coast Rate m/yr. m/yr. m/yr. m/yr. Average -0.02 -0.04 -0.001 -0.008 change net Average Peros change 0.18 0.18 0.16 0.24 Banhos expansion Average -0.24 -0.34 -0.15 -0.28 change retreat Entire Normal Average -0.07 -0.06 Coastline1 Coastline2 ̶ change net -0.08 -0.02 Diego Average Garcia change 0.25 0.21 0.27 0.16 ̶ expansion Average -0.33 -0.25 -0.37 -0.17 ̶ change retreat

Table 2 reveals that the ocean-facing coastlines of the Peros Banhos islands displays the highest rates of retreat. Analysis of individual islands, however, shows that not all ocean-facing coastlines on this atoll have retreated in the considered time period. For instance, sectors of the ocean-facing coastline of several islands distributed on the western and northern rim of the atoll show pronounced expansion (Fig. 6A, B, C, and D).

Examination of the time-separated remote sensing imagery reveals that these areas of coastline expansion have been delivered by the beachward migration of the brown-soil line, presumably driven by the conversion of beach sands to vegetated island interior. This mechanism, which has expanded islands by as much as 15–20 m in the last half-century, is particularly important on the islands which ornament the western rim of Peros Banhos, such as Grande Mapou (Fig. 6A). For islands on the northern rim of this atoll, this

mechanism of island growth is also present, but restricted to ocean-facing embayments, such as exist on Ile Manoel (Fig. 6B) and Ile Diamant (Fig. 6C). Ile Diamant shows an additional mechanism of growth. Here, the area behind a promontory on the ocean-facing side of the island which existed in 1979 has completely infilled by 2015, serving to expand the island by >30 m (white arrow in Fig. 6C). By contrast to the dynamic ocean-facing coastlines of Peros Banhos, the lagoon-facing coastlines are all very stable. If change can be detected, however, it tends to be a modest retreat (such as observed for Ile Longue, Fig.

6D). This lethargy is lacking for the hoa-facing coastlines of the Peros Banhos islands which are particularly dynamic. For instance, the tips of the islands of both Petit Soeur (Fig.

6E) and Ile Diamant (Fig. 6F) both retreated by more than 30 m in the period of observation.

Further, the rapid accretion of island tips caused the hoa-facing coastlines of other atolls to expand considerably – particularly so for highly elongate islands such as Grande and Petite

Soeur (Fig. 6G). Tips can also radically decrease in size through time (e.g. Fig. 6A - Grande

Mapou and Fig. 6H - Grande Soeur).

Switching attention to Diego Garcia, local examples of both expansion and retreat of the ocean- and lagoon-facing coastlines alike can be identified in the time-separated remote sensing data (Fig. 7). Although the filling of embayments is a rare motif of coastline advance for the Peros Banhos islands, it is a common mechanism of growth on the lagoon- facing side of Diego Garcia. As can be illustrated with the remote sensing imagery, many of the cuspate embayments which perforate the lagoon-facing margin of this island have infilled between 1963 and 2013 (Fig. 7A, B, and C), suggesting this mechanism to be important for the evolution of Diego Garcia, as captured in Table 2 which emphasizes how the annualized rate of coastline migration decreases markedly for Diego Garcia if its

southern embayments are not included in the analysis (0.08 vs. 0.02 m/yr.). When land is gained on the lagoon-facing side of Diego Garcia, it is typically balanced by land lost on the ocean-facing side, as emphasized in the vicinity of Rambler Bay (Fig. 7D). Similarly,

Barton Point (Fig. 7E), situated in the north of the atoll, has lost approximately 15 m of its ocean-facing coastline in the last 50 years (Fig. 7E), balanced by an equivalent extension of the lagoon-facing coastline. The western limb of the island in the south (Fig. 7F) and narrow eastern limb of the mid-island (Fig. 7G) also show the common pattern of land lost from the ocean-facing coastline and minor gains on the lagoon-facing side. As always, however, there are exceptions to the rule with pronounced retreat recorded on both ocean- and lagoon-facing coastlines, as documented around Simpson Point (Fig. 7H).

Figure 6. Coastline-shift maps highlighting areas of land gained and lost between 1979 and 2015 for the islands of Peros Banhos. Hot colors denote areas of coastline retreat and cool colors denote expansion. Grayscale WorldView-2 satellite image acquired in 2015 is used as base layer. (A), (B), (C) and (D) illustrate that the ocean-facing coastline of several islands distributed on the western and northern rim show pronounced expansion, which has been delivered by the beachward migration of the brown-soil line. (A) shows the expansion of entire ocean-facing coastline, (B) and (C) are taken from northern rim, where pronouncing expansion could be mainly found on the ‘U’-shaped part of the coastline. (C) further shows a second mechanism for the seaward expansion of an island-embayment infilling. (D) emphasizes the overarching pattern of coastline migration in Peros Banhos whereby the ocean-facing coastline expands while the lagoon-facing coastline retreats. Meanwhile, the hoa-facing coastlines are particularly dynamic. The tips of the islands in (E), (F) both retreated by more than 30 m. (G) highlights the rapid accretion of sand spits. Both (A) and (H) exhibit marked erosion of land tips.

Figure 7. Coastline-shift maps highlighting areas of land gain (blue) and loss (red) between 1963 and 2013 in Diego Garcia. High-resolution satellite imagery acquired in 2013 is used as the base layer. (A) through (C) illustrates how tranquil embayments on the lagoon-facing coastline of the island are highly dynamic and swiftly infill through time, thereby increasing the area of the island. Cases of land lost on Diego Garcia tend to be on ocean- facing coastlines (D, E, and F). In rare cases, however, ocean-facing coastlines do prograde seaward, such as resolved in (H) and (I), accompanied by a loss of land on the lagoon side. Seaward expansion of Simpson Point (H) is particularly pronounced.

2.4.2 Coastline Dynamics Very with Island Shape and Size

Trends between the shape of the Peros Banhos islands and the behavior of their coastlines through time were explored by computing the principal axes ratio (PAX) of each island using the workflow described by Purkis et al. (2007). These data were considered in unison with the size of the islands (Fig. 8). Equivalent analysis cannot be conducted for

Diego Garcia since it hosts a single large island which inhabits the majority of the rim of the atoll. Of the 20 Peros Banhos islands for which we have time-separated remote sensing data, exactly half display an increase in island area over the 36 years of observation (Fig.

8A), and ten show a decrease in size (Fig. 8B). To the former, no systematic behavior is seen between the size of an island and its coastline dynamics. More circular islands, however, tend to have limited expansion, whereas those which are more elongate are more dynamic. The same trend is seen for the islands that have reduced in size (Fig. 8B).

2.4.3 Spatial Patterns of Island Alteration in Peros Banhos and Diego Garcia

Because the atoll-island coastlines are fronted by different hydrodynamic conditions, it is reasonable to assume that they will undergo variable change according to whether they face the open ocean or the lagoon. To explore this premise, the terrestrial area gained or lost over the 36 years of observation was quantified for each of the 20 Peros Banhos islands for which time-separated remote sensing data exist (Fig. 9). A distinction was made between those islands which were less than 20 ha in area at the time the atoll was imaged in 1979, and those which were larger than this value. Data were assembled to consider the performance of the complete island coastlines, agnostic to whether they are ocean-, lagoon, and hoa-facing (Fig. 9B), as well as each of these three categories in isolation (Fig. 9C, D, and E, respectively).

When the entire coastlines of the islands are considered in aggregate (Fig. 9B), it is revealed that the islands situated in the north of Peros Banhos atoll are all increasing in size, save for Ile de la Passe (#12 in the figure) whose coastline is retreating. By contrast, all of the islands situated on the west of the atoll lost area between 1979 and 2015. If only the ocean-facing island coastlines are considered (Fig. 9C), 15 out of 20 islands, regardless of their position on the atoll rim, have gained in area. The five exceptions to this trend are

Ile Pierre (#4 in the figure), Ile Poule (#11), Ile de la Passe (#12), Mapou Du Coin (#19) and Ile Fouquet (#20). The opposite trend is observed for the lagoon-facing coastlines (Fig.

9D), the majority of which have lost area. Similarly, the hoa-facing coastlines also tend to decrease in area, regardless of position on the atoll (Fig. 9E).

In order to understand possible drivers of the different behavior of ocean- versus lagoon-facing coastlines, hydrodynamic exposure across the archipelago was calculated on

the basis of fetch length convolved with wind speed and direction. The hoa-facing coastlines were not considered in the equivalent analysis for Diego Garcia, as Diego Garcia hosts a single large island. Therefore, this category of coastline is not applied to this atoll. .

As captured by the meteorological station at the Diego Garcia airfield (from 2004 to 2015), the Chagos Archipelago is situated in the trade-wind belt from May to September during which time the wind blows strongly from the southeast. Low velocity cyclonic winds characterize the rest of the year (Fig. 10A). The calculation delivers an estimation of the magnitude of wave energy in units of joules per m3 for Peros Banhos (Fig. 10) and for

Diego Garcia (Fig. 11). To Peros Banhos first - the modelled hydrodynamic setting suggests that for the west side of this atoll, the trade winds deliver higher energies on the lagoon-facing island coastlines than those which are ocean facing. Here the hoa hosts a very high hydrodynamic energy regime, which allow sediment fluxes from ocean to lagoon and thus influence nearby coastline change. Meanwhile in the northwest of Peros Banhos, energies on the lagoon-facing islands are also higher than on the ocean facing, but, because the hoas are aligned parallel with axis of the trade winds, they are particularly high energy.

When interpreting these data, it must be remembered that the coastlines experience prevailing wave energy after it has been attenuated by the reef flats which fringe each island and steep fore reef (Sheppard et al., 2005; Quataert et al., 2015). It is therefore imperative to also consider the width of the reef flat which was quantified from satellite imagery. Here, there is a negative relationship (R2 = 0.37) between the magnitude of coastline change

(1979-2015) for the 20 Peros Banhos islands, reef-flat width, and prevailing wave energy

(Fig. 10C). This positive relationship, albeit weak, suggests that more energetic coastlines tend to be fronted by wide reef flats and retreat through time. Also emphasized is that the

ocean-facing coastlines of the Peros Banhos islands tend to expand through time, while the lagoon-facing retreat, as previously confirmed by time-separated remote sensing (Fig. 9).

The map of simulated wave energy for Diego Garcia (Fig. 11) has similarities, but also important differences, to that produced for Peros Banhos (Fig. 10). In terms of similarities, the prevailing wave energy in the Chagos, which is driven by the trade winds, delivers the maximum hydrodynamic energy to the southeast-facing coastline of the atoll island, as it does for the islands of Peros Banhos. The difference is that for Peros Banhos, the most energetic coastlines are lagoon-facing as the atoll’s protective reef rim is most poorly developed in its southeastern quadrant (see Fig. 1B). Meanwhile for Diego Garcia, which has a well-developed reef rim around the entirety of the atoll, save for a small hoa at its northern tip (the leeward side of the atoll), the lagoon is modelled to be quiescent, receiving zero impact from prevailing long-period ocean swell, which is instead concentrated on the ocean-facing coastlines of the island, and those which face southeast, in particular. Configured as such, it might be anticipated that the southeast-facing oceanward coastline of Diego Garcia would be particularly vulnerable to erosion. For instance, work in French Polynesia by Le Cozannet et al. (2013) suggests that temperature- induced coral bleaching serves to stifle sediment supply to atoll islands. Indeed, Purkis et al (2016) results showed that this coastline has retreated during the period of observation, but only modestly so (see their Fig. 9). This behavior for Diego Garcia is in contrast to that of Peros Banhos, where there is no clear relationship between wave energy and coastline dynamics, regardless of the width of the island-sheltering reef flat (Fig. 11B).

Figure 10. Hydrodynamic energy as a driver of coastline change of the 20 Peros Banhos islands. (A) A pair of seasonal wind roses capturing prevailing wind speed and direction from the naval air station on Diego Garcia for the period 2004 through 2015. (B) Map of modelled seasonal local wave energy incident on the ocean- and lagoon-facing coastlines of the Peros Banhos. Insets emphasize that for the west of the atoll, the trade winds deliver higher energies on the lagoon-facing island coastlines than those which are ocean facing. (C) Plots coastline migration for the period 1979-2015 versus reef-flat width. The wave energy incident on each coastline is denoted in color and circles denote ocean-facing coastlines and squares lagoon-facing. More energetic coastlines are fronted by wide reef flats and retreat through time.

Figure 11. The hydrodynamic setting of Diego Garcia. (A) Map of modelled wave energy incident on the ocean-facing coastlines of the island. (B) Plots of coastline migration for the period 1963-2013 (y-axis) versus reef-flat width (x-axis). The wave energy incident on each coastline is denoted in color and circles denote ocean-facing coastlines. Unlike the case for Peros Banhos, there is no clear relationship between the hydrodynamic energy incident on the coastline of Diego Garcia, the width of the reef flat, and the change in island size through time.

2.5 Discussion

Pairing the two sites in the same archipelago allows the dynamics of atoll islands to be contrasted while avoiding bias introduced by varying ocean climate. The atolls of Diego

Garcia and Peros Banhos are geomorphologically different and serve to deliver divergent hydrodynamic environments, which, in turn, delivers dissimilar styles of sediment transport, and therefore different coastline dynamics. However, at a high level, the behavior of the islands on the two atolls is similar in the sense that island area is almost static through the half-century of observation (Table 1). In neither case, though, are the coastlines of these islands immobile. Instead, they are highly dynamic, but the island area lost is almost exactly balanced by area gained, during the period of observation. These similarities, however, mask pronounced differences in the style of island dynamics between Peros

Banhos and Diego Garcia. For the former, island expansion is mediated by extension of the ocean-facing coastlines and retreat by the erosion of the lagoon-facing coastlines. This pattern is opposite to that of Diego Garcia, which sees substantial land gained on the lagoon-facing side of the island and lost on the ocean facing.

Models of hydrodynamic exposure aid in the explanation of these trends. Despite occupying the platform interior, the Peros Banhos lagoon is far from tranquil because the rim of this atoll is not competent in its southeastern quadrant, allowing long-period ocean swell to enter the lagoon, driven by the southeasterly trade winds which prevail from May through September (Fig. 10). Fronting this hydrodynamic energy, the lagoon-facing coastlines of the Peros Banhos islands retreat, whereas the ocean-facing ones, situated in a relative lee during the trade-wind season, expand, as sediment is hydrodynamically moved down wind. Despite being a smaller atoll than Peros Banhos, the geomorphology of Diego

Garcia also differs substantially. Diego Garcia is unusual in the length and continuity of its reef rim, which extends round 90 percent of the atoll circumference, save only for a limited aperture on its northern face, as originally reported by Stoddart (1971). This configuration delivers a quiescent lagoon environment (Fig. 11), and accordingly, the majority of coastline extension is witnessed on the tranquil lagoon-facing coastline of this island (Fig.

7). This behavior is consistent with the observations of Rankey (2011) who noted lagoon- facing coastlines were more dynamic than ocean-facing ones for some atolls (Maiana,

Aranuka) in the Gilbert Islands (Kiribati). The reason for this difference being that the lagoon-facing coastlines lacked the stabilization of well-developed fringing reefs, as present on the ocean-facing coastlines of the islands, as is the case for Diego Garcia.

Isolated from the sediment produced in its substantial lagoon by the island-hosting rim, the majority of the ocean-facing coastline of Diego Garcia is retreating.

2.5.1 Attempt for Unraveling Anthropogenic Influence on Island Behavior

By comparing Peros Banhos, which is unsettled by humans, to Diego Garcia, which hosts between 1,000 and 5,000 U.S. troops and civilian support staff, there is the opportunity to gain insight into anthropogenic influence on atoll-island dynamics. For

Diego Garcia, Purkis et al. (2016) had previously examined the within-atoll influence of humans on coastline migration. The authors concluded that the areas suffering the most acute erosion correlated with those with the highest density of infrastructure, including areas where the coastline has been armored with concrete, etc., such as has been implemented at Simpson Point, Point Marianne, and Eclipse Point.

To test the hypothesis that coastlines situated adjacent to densely-settled parts of atoll islands undergo accelerated coastline dynamics, as compared to natural atoll islands, the

rates of change were compared for the ocean-facing coastlines of Peros Banhos to those of

Diego Garcia. Three categories of coastline were considered between the two atolls: stretches far from human modification on both Diego Garcia [1] and Peros Banhos [2]; then [3], stretches of coastline in the immediate vicinity of human modification (on Diego

Garcia only). Since reef flats serve to attenuate wave energy, the stretches of coastline assembled in scenarios [1], [2], and [3] were all fronted by reef flats with widths varying between 160 m and 200 m. These criteria resulted in the selection of 5 km of the ocean- facing coastline of Diego Garcia in Category [1], 6 km of ocean-facing Peros Banhos coastline in Category [2], and 6 km of Diego Garcia coastline in Category [3] (Fig. 12A).

Coastline shifts per decade were audited along nodes spaced by 20 m for the three categories and the average and variance of the migrations plotted (Fig. 12B). The data reveal that the stretches of coastline offset from human impacts in both Diego Garcia and

Peros Banhos (Categories [1] & [2]) have tended to expand at an average rate of 1.0 m per decade. By contrast, the stretches of coastline in the vicinity of human disturbance on

Diego Garcia (Category [3]) have retreated at a rate of approximately -2 m per decade. The behaviors of the unimpacted verses impacted coastlines are statistically different (p = 0.05), suggesting that human disturbance has the capability to accelerate coastline erosion. Ford

(2012) ascribed similarly enhanced erosion of disturbed stretches of atoll-island coastlines in the Pacific to reduced rates of calcification of the offshore reefal and foraminiferal assemblages, likely caused by increased turbidity and nutrient loading imposed by human activity. Duvat (2020) highlighted that the ‘reclamation-fortification island model’ in the

Maldives Archipelago would compromise the capacity of the reef-island system to naturally adjust to sea-level rise in the future. This limitation would highly depend on the

specific technical requirements (such as infrastructure maintenance and upgrade) and huge financial investments. Note that we could not conduct an equivalent test for the lagoon- facing coastlines of the atolls because fringing reefs are virtually absent from the interior of Diego Garcia, though wide and well-developed in Peros Banhos, thereby preventing a like-for-like comparison of dynamics.

The behavior of the Chagos islands can be placed in a global context via comparison with equivalent studies which capture coastline dynamics for select islands in French

Polynesia (Le Cozannet et al., 2013; Yates et al., 2013; Duvat & Pillet, 2017; Duvat et al.,

2017), Tuvalu (Kench et al., 2015; 2018), the Marshall Islands (Collen et al., 2009; Ford,

2012; Ford, 2013; Ford and Kench, 2015), Kiribati (Biribo and Woodroffe, 2013; Rankey,

2011), and the Federated States of Micronesia (Webb and Kench, 2010) and one study from the Maldives in the Indian Ocean (Aslam & Kench, 2017). A common thread through the majority of these studies is that per-decade changes in island area equivalent to 3% of the total island area, or less, are considered inconsequential, which is a threshold also adopted for this comparison (Fig. 13 – islands with areas that change <3% per decade plot between the two horizontal dashed blue lines). To aid the comparison, a threshold in total island size of 20 ha (vertical dashed black line in Fig. 13) was useful for distinguishing differences in island dynamics. For the 559 islands that are < 20 ha in area, 29% exhibit a decadal change exceeding 3% of their total area. For the >20 ha islands, of which there are 171, only 12% exceed this threshold in dynamics. As emphasized in Figure 13, adding the Chagos to this database shows its islands to perform within the bounds of those from the nearby Maldives

Archipelago. Such consistency between ocean basins is surprising given the varied oceanographic forcings, differences in diversity of sediment producers – including disturbances to those producers in terms of coral bleaching, etc., and, perhaps most intriguingly, differences in frequency of impacts from storms and cyclones (Perry et al.,

2011; Le Cozannet et al., 2013; Yates et al., 2013; Duvat et al., 2017). This consistency in behavior carries at least two important lessons. First, it reaffirms that the coastlines of small islands are particularly dynamic, though as emphasized in Figure 13, the probability of perceptible erosion is approximately equal to the chance of island growth. Second, there

seems to be universal bounds to the extent of coastline dynamics, regardless of the physical and environmental setting of the atoll islands.

There are many reasons why rate of sea-level rise might impart control on atoll-island behavior, but two stand out. The first reason is that high rates of rise would be anticipated to amplify the hydrodynamic energy incident on coastlines, particularly so during storms, thereby increasing their erosion (Roy and Connell, 1991; Dickinson, 1999; Woodroffe,

2008; Church et al., 2006; Nicholls and Cazenave, 2010). Second, rapid sea-level rise decreases the ability of key carbonate producers to create sediment by virtue of the link between solar irradiance and production rates (Schlager, 1993; Van Woesik and

Cacciapaglia, 2018) and islands starved of sediment are also more dynamic (Perry et al.,

2011; Morgan & Kench, 2016). Sea-level control on island behavior is not upheld by data.

Fig. 14A captures global rates of sea level rise averaged over the period 1993 to 2018 as provided by the French Archiving, Validation and Interpretation of Satellite

Oceanographic data program (AVISO) using satellite altimetry (T/P and Jason1/2/3, plus available ERS-1/2, Saral and Envisat). The atoll islands in the Indian and Pacific Oceans for which there exists data on coastline behavior span a broad range in rates of sea level rise (Fig. 14B). For instance, Tuvalu and the Marshall Islands are experiencing double the rate of rise experienced in French Polynesia (4 mm/yr. vs. 2 mm/yr., respectively). Decadal change in atoll-island landmass is, however, uncorrelated with rate of sea-level rise, suggesting that local factors are the primary determinants of these dynamics, at least for the seven atoll systems (including 671 islands) assembled in Figure 14.

Chapter 3: Low-Energy Facies Dynamics in Andros Island,

Bahamas

3.1 Background

The tidal flats flanking the west coast of Andros Island (Great Bahama Bank) are composed of a complex suite of sub-environments. In particular, the flats are characterized by dendritic networks of tidal creeks which incise through a substantial Holocene lime- mud wedge in the intertidal zone and a large mat-covered supratidal zone. Since the Andros tidal flats span the intertidal and supratidal zones, their morphology is sensitive to even modest sea level change (Rankey and Morgan, 2002). Perplexing problems confronting sedimentologists are an understanding of tidal flat evolution, their relative importance in the stratigraphic record, and petrophysical attributes which permit these ‘low-energy’ sediments to develop into potential seal facies or occasionally into hydrocarbon reservoirs

(Roehl, 1967; Fagherazzi et al., 2006; Maloof and Grotzinger, 2012; Jahnert and Collins,

2013). Since the late 1960s, studies have extensively assessed patterns of both modern and ancient carbonate tidal flats and provided insights into their lithologic attributes (e.g. Shinn et al., 1965; Laporte, 1967; Roehl, 1967; Goldberg, 1967; Shinn et al., 1969; Hardie, 1977;

Rankey and Morgan, 2002; Rankey, 2002; Rankey et al., 2004; Maloof and Grotzinger,

2012). Several earlier studies focus on the sedimentology or stratigraphy of modern carbonate tidal flats and their ancient analogs. For example, Shinn et al. (1965) recognized that pene-contemporary or early diagenetic dolomite occurs on the Andros tidal flats, and

Shinn et al., (1969) described sub-environments and facies of the Andros tidal flat complex in detail including their distributions, sedimentary structures, sediments, living fauna and

of the tidal-flat complex is transgressive in nature, whereas the flats to the southwest were prograding. Rankey and Morgan (2002) characterized the spatial patterns of sedimentation and analyzed the morphology in the Triple Goose Creek of the Andros tidal flats. Rankey et al. (2004) found no substantial impact on facies patterns caused by Hurricane Michelle, and they suggested that the geologic evolution and preservation of ancient tidal-flat successions may have been shaped by mega-storms that simply cannot form in the present global climate. Maloof and Grotzinger (2012) provided a model for the evolution of the

Andros tidal flats based upon facies distribution from coring and an analysis of the external

(e.g. sea level) and internal forcing (e.g. migrating channels) on the modern tidal facies mosaic.

These studies provided an understanding of facies successions in the context of historical sea level change. Here, we build forward from these studies and use time- separated remote sensing data spanning 70+ yrs. and quantitative insights that combine lateral and vertical facies dynamics and their relationships in order to better understand the evolution of the Andros tidal-flat system.

Adapting the scheme proposed by Shinn et al. (1969) and Hardie (1977), we partition the Andros tidal-flat belt into three shore-parallel zones: (1) the subtidal platform interior;

(2) the intertidal-channelized zone, and (3) the supratidal inland Scytonema marsh (Fig.

15). The latter zone is named by virtue of the extensive colonization by Scytonema, a genus of multicellular filamentous cyanobacteria which forms laterally-continuous mats. The presence of Scytonema mats is not restricted to the supratidal zone but laterally discontinuous, discrete patches of Scytonema also inhabit areas within the intertidal

diagenetic features. Based on their analysis, they concluded that the northwestern portion channelized zone that are of sufficient elevation to be supratidal. Because of their dark color, these patches are easily resolved on remote sensing data.

Figure 15. Schematic cross-section of the Andros tidal flats modified from Hardie (1977) showing the three zones into which the deposit can be conveniently divided. From seaward to island-ward, these are the subtidal platform interior, intertidal-channelized zone, and supratidal Scytonema marsh. The overall goal of this study is to understand the sedimentological response of carbonate depositional systems to changing sea level. To achieve this goal, and in the context of sea-level oscillation over the last half century, the specific aims of this study focus on the intertidal-channelized zone and: (i) examine the migration of the seaward and landward boundaries of the zone, (ii) reconstruct the dynamics of patterning of the

Scytonema patches, (iii) quantify the evolution of the length of the tidal network of creeks which transverse the intertidal channelized zone, and (iv) assemble these data to predict how the very earliest onset of a transgression might be recorded in the rock record of a tidal-flat environment. The study builds upon a growing body of literature (e.g. Hardie,

1977; Shinn et al., 1965; Shinn et al., 1969; Rankey & Morgan, 2002; Rankey et al., 2004;

Maloof & Grotzinger, 2012) which suggests that the evolution of carbonate tidal flats is predominantly determined by allogenic drivers – sea level, in particular – which is in stark contrast to the radical and short-term reconfiguration of equivalent siliciclastic settings, wrought by autogenic drivers within the system (Fagherazzi, 2008).

To achieve the four specific aims, the study calls upon data collected in the field in

Triple Goose Creek, sea-level data recorded at the three most proximal research-quality tidal gauges to Andros Island (Settlement Point, Grand Bahama Island; and Virginia Key and Key West, Florida), historical storm tracks across the Bahamas, aerial photographs acquired in 1943, and a portfolio of high-resolution satellite imagery spanning 1985 through 2018 assembled in Google Earth Engine.

3.2 Geological and Environmental Setting

3.2.1 Geological Setting

The Great Bahama Bank (GBB) is located at the southern extremity of North

America’s eastern continental margin and is the largest isolated carbonate platform on

Earth (Maloof & Grotzinger, 2012; Harris et al., 2015). The GBB is a modern location of major carbonate deposition, which forms a fundamental understanding of many depositional processes of modern carbonate sedimentation. The GBB serves as a training venue for academia and industry, is the basis for numerous geological models, and is often used as a reservoir analog (Harris et al., 2015; Purkis & Harris, 2016). Andros Island, the largest island at 5,957 km2 in size, is located on an eastern facing margin of the western portion of the GBB (Fig. 16A). Andros is bounded by the seaways of Tongue of the Ocean

(TOTO) immediately to the east and the Florida Straits to the west. The coastlines of

Andros Island extend along a northwest-southeast axis. On the eastern side of Andros,

Pleistocene eolian ridges run parallel along the strike of the island. The western side of the island is low and flat, with Holocene mudstones forming a shore-parallel belt which is more than 120 km along the coastline – the Andros tidal flats (Shinn et al., 1969). The Holocene mud wedge is overlaying a Pleistocene ooid shoal complex that was likely deposited during the last interglacial. Cores and ground penetrating data document a tidally formed shoal system that has sediementary characteristics similar to the modern Joulter Cays ooid shoal that is now located north of Andros Island (Hazard et al., 2017).

3.2.1 Environmental Setting

The climate of Andros Island is typically tropical-maritime with long, warm humid summers (May to October) and mild, dry winters (Hardie, 1977). The annual average temperature is 25 ℃ and the annual rainfall can range from 232 cm - 65 cm, with an average of 129 cm per year (Hardie, 1977; Maloof and Grotzinger, 2012). Wind data was retrieved from Harris et al. (2015), where QuikSCAT satellite observations provided wind speed and direction for the period January 2013 to January 2014. The direction of trade winds originates primarily from the east (Fig. 16B). Trade winds are strongest and most dynamic during March and April. Throughout a period of 2.5 years, the tide was found to be semidiurnal, with Andros experiencing two high tides and two low tides each day

(Ginsburg and Hardie, 1975). The tidal height measured by Hardie (1977) in the channel mouth and ponds is 46 cm and 29 cm, respectively. As illustrated by Hardie (1977), tidal heights are controlled by weather patterns such as strong winds, air pressure and rainfall

(especially). The height of tides determines whether sub-facies will be flooded or exposed.

Since the daily tide is strongly dependent on day to day weather, the variability of local weather increases the uncertainty of tidal height prediction (Hardie, 1977).

The Bahamas in situated in the main track of West Indian tropical cyclones. Andros

Island is hit by a hurricane on an average of once every 2.5 years usually between June to

October (Hardie, 1977). Notable hurricanes in the modern era have included Hurricanes

Betsy (1965), David (1979), Arlene (1987), Andrew (1992), Lili (1996), Floyd (1999),

Michelle (2001), Wilma (2005), and Matthew (2016). Three major hurricanes which passed directly over Andros Island were included in the analysis of this study spanning from 1985-2018: two Category 5 hurricanes - Andrew (1992) & Matthew (2016) and one

Category 4 - Michelle (2001). These three hurricanes, and their meteorological history, are described below (Fig. 16C). First, hurricane Andrew (pink track in Fig. 16C) was the strongest tropical cyclone of the 1992 Atlantic hurricane season and lasted from mid to late

August. Hurricane Andrew developed from a westward tropical wave which moved off the coast of Africa on August 14th and peaked as a Category 5 hurricane on the Saffir–Simpson hurricane wind scale while passing directly over the north dip of the Andros Island

(Rappaport, 1993). The second hurricane is Michelle (yellow track in Fig. 16C) which was the fifth costliest tropical cyclone in Cuban history. Hurricane Michelle was the strongest of the 2001 Atlantic hurricane season. It developed from a tropical wave that traversed into the western Caribbean Sea on October 29th. Michelle began to accelerate northeastward and reached its peak intensity as a Category 4 hurricane with winds of 140 mph (225 km/h)

(Beven, 2002). Lastly, hurricane Matthew (green track in Fig. 16C) is presently the last

Category 5 Atlantic hurricane that passed through Andros since 2007. Matthew originated from a tropical wave and developed into tropical storm in the eastern Lesser Antilles on

September 28, 2016. After several landfalls in Haiti and Cuba, it weakened somewhat but re-intensified as it tracked northwest. Lastly, hurricane Matthew made landfall in the

northern Bahamas in October 5th. The storm then travelled parallel to the southeastern coast of Florida over the next 36 hours (Stewart, 2017).

Figure 16.(A) The Great Bahama Bank (GBB) is the largest isolated carbonate platform on Earth and situated at the southern extremity of North America’s eastern continental margin. Andros Island in the red box is the largest island atop GBB, which located along a windward margin of the platform. (B) Shows the wind rose capturing wind speed and direction extracted from QuikSCAT satellite observations for the period January 2013 to January 2014. (C) Three major hurricanes passed in close proximity to Andros between 1985 and 2018: Hurricane Andrew (Category 5) in 1992, Michelle (Category 4) in 2001, and Matthew (Category 5) in 2016, and are demarked in pink, yellow and green, respectively. 3.3 Material and Methods

3.3.1 Pre-Processing of Tide Gauge Data

The recent sea-level history for the Bahamas was reconstructed for the period 1985-

2018 from an array of three proximal (200-300 km) tidal gauges (Fig. 17A-C) and processed into annual averages (dots in Fig. 17D). Although the measured daily tidal range could reach up to +100 cm, the raw monthly tide gauge data were processed to eliminate

the daily and seasonal fluctuations due to coastal ocean temperatures, salinities, winds, atmospheric pressures, and ocean currents. A smoothed sea-level trend was also reconstructed from the raw data via application of a 3-yr. moving average (thick lines in

Fig. 17D). In addition, the lunar nodal cycle (LNC) of the average of the three sea-level curves was computed. The LNC is produced by the varying declination of the Moon over a period of 18.61 yrs. and drives changes in tidal amplitude globally (Baart et al., 2012).

Figure 17. Sea-level elevations recorded from 1985-2018 for three tide gauge stations Key West, Virginia Key and Settlement Point. (A) shows the locations of the three tide gauge stations. Their sites are close to Andros Island where studies have been undertaken. (B-D) shows monthly raw sea level fluctuations in the period 1985-2018. Monthly raw data was then processed into annual averages (dots) and 3-yr. moving average trend (thick lines) in (E). Note how the 18.61 yr. lunar nodal cycle (LNC)-gray line in (E) serves to amplify sea- level rise when in phase with the sea-level fluctuations induced by other factors, but dampen it when out of phase (denoted by white arrows). As the white arrows shown in Figure 17, the LNC is important as amplifies sea-level rise during its ascending phase but dampens the rise in its descending phase. These effects can exceed 2 cm of relative sea level, which is substantial considering that the linear relative sea-level trend measured in the Settlement Point was between 0 and 5 mm yr-1 with

95% confidence interval (data provided by tidesandcurrents.noaa.gov)

3.3.2 Pre-Processing of Vintage Aerial Photographs

As illustrated in Purkis et al. (2016), the negatives of the vintage aerials were scanned at 600 pixels-per-cm to yield digital images with an effective spatial resolution of 1.0 m.

The shortcoming of these vintage photographs, however, is that they do not contain any geo-information and have to be corrected via georectification, a transformation process used to project an unpositioned (or poorly positioned) historical image onto a known coordinate system. As explained in chapter 2.3.1, this procedure involves pairing the archive data with a well-positioned satellite image and selecting points on the ground common to both. These locations become reference points in the subsequent warping of the unpositioned image onto a coordinate system. This task was performed in ArcGIS 10.7 through selecting easily recognized ground control points in both the vintage aerial photographs and modern (2018) WorldView-2 satellite imagery, which were presumed to have temporal stability, such as blue holes, main channels and ponds. The error generated by georectification was less than 2m.

3.3.3 Pre-Processing of Satellite Images

The satellite imagery assembled for Andros Island were selected every three years for the period 1985 to 2018 within Google Earth Timelapse (data provided by earthengine.google.com/timlapse.com). Google Earth Timelapse is comprised of 35 cloud- free annual global mosaics from 1984 to 2018 gathered by Carnegie Mellon University

CREATE Lab’s Time Machine library, which is a technology for creating and viewing zoomable and pannable timelapses over space and time. The majority of the satellite images come from Landsat, a sensor which was first lofted in 1972. In 2015-2018, imageries from Landsat 8 and Sentinel-2A (part of the European Space Agency’s

Copernicus Earth Observation Program) were combined to deliver high-resolution optical imagery which excellently complements data continuity as provided by SPOT- and

Landsat-type data. The resolution of satellite imagery captured in 1980s-2000s ranges from

5-10 m/pixel. As a result of the variation of imagery quality, the geo-positioning of the modern satellite representation was checked against the 2018 WorldView-2 satellite imagery using ArcGIS (10.7) and deemed to be correct.

3.3.4 Change Detection from Time-Separated Remote Sensing Data in the Andros Tidal Flats

As defined by Shinn et al. (1969) and Hardie (1977), the distal boundary of the intertidal-channelized zone of the Andros tidal flats is demarked by the transition into the subtidal platform interior, i.e. the coastline. This zone’s proximal boundary, by contrast, is represented by the transition into the supratidal inland Scytonema marsh. As developed in

Figure 15, both boundaries are spectrally distinct and could be reliably digitized in the portfolio of satellite imagery assembled for this study. Because the boundaries of the intertidal-channelized zone are particularly crisp within the remote sensing data, this analysis was conducted in four 280 sq. km focus areas distributed down the western margin of Andros Island. From north-to-south, these focus areas are Triple Goose Creek, Blue

Creek, Middle Bight, and Cormorant Point (Fig. 18). Triple Goose Creek represents a rare

Holocene carbonate depositional system with the well-developed channelized zone that may have been more globally abundant during much of Earth history (Maloof and

Grotzinger, 2012). The Triple Goose Creek area has been examined by sedimentologists since the 1960s (Shinn et al., 1965; Shinn et al., 1969). For this reason, Triple Goose Creek was targeted so as build forward from previous studies (e.g. Rankey and Morgan, 2002).

The second of our focus areas, Blue Creek, was selected because it hosts a particularly well-developed and expansive (11 km) intertidal channelized zone. Further, the Scytonema patches in the intertidal channelized zone are excellently resolved throughout the Google

Earth Timelapse. The third and fourth focus areas, Middle Bight and Cormorant Point, respectively, are located on the southwest coast of Andros Island and have coastlines which are orientated perpendicular to those of the other two focus areas, thereby providing contrasting hydrodynamic conditions.

Changes in the position of the seaward and landward boundaries of the intertidal- channelized zone were quantified from the period of 1985 through 2018 in the four focus areas using the Digital Shoreline Analysis System (DSAS), an extension to ArcGIS (Fig.

19A-B). DSAS detects changes by recording the intersection of shorelines and transects

(with 30m interval in the study) cast perpendicular to a user-generated baseline (Fig. 19A)

(Thieler et al., 2009; Ford and Kench, 2015). Next, statistics were then calculated automatically according to the position of the transects (Ford and Kench, 2015). Transects were colored to represent the distance of migration in meters (Fig. 19B). Within the intertidal-channelized zone, changes of the patterning of Scytonema patches were manually digitized in the time-separated remote sensing data within the same focus areas. In the study, the annualized changes of area and number of the patches were assessed. Dynamics of all these Scytonema-dominated facies were considered in the context of recent oscillations of relative sea level (Fig. 17).

Figure 18. Analysis of the migration of the boundaries of the intertidal- channelized zone was conducted in four focus areas distributed from North to South along the western coastline of Andros Island. The focus areas capture variation within the channelized zone and as well as varying orientation of the coastline with respect to prevailing hydrodynamic energy.

3.3.5 Quantifying Change in the Configuration of the Andros Tidal-Channel Network

Through Time

Adapting the workflow of Harris et al., (2011), we captured the network of tidal channels which incise the Holocene Andros tidal flats by developing a skeleton of vectors which capture the principal axes of the channels. Each element of the skeleton corresponded to the midline of each and every channel which could be resolved from the high-resolution remote sensing data. In this way, the midline was defined as having equal distance from both sides of the considered channel. Channel skeletons were generated for the tidal network as resolved in 1943 from the vintage aerial photographs and from 2018, as imaged in the modern satellite imagery. The vintage and modern channel skeletons were quantitatively compared to detect changes in the overall length of the channel networks through time, and to identify cases where channels have been created and abandoned during the period of observation.

Data were manually extracted from the vintage aerial photographs (1943), and the modern satellite imagery (2018) was assembled to represent the length of the tidal network in the four focused areas at the Triple Goose Creek (Black boxes in Fig. 20A). Each element of the channel skeleton was attributed one of five orders on the basis of the number of bifurcations separating the considered channel from the ‘main’ channel, defined as connecting the overall network to the coastline of the tidal flats (Fig. 20B-E). This construct delivered a channel-bifurcation hierarchy, whereby the main channel was designated as

‘Order 1’ (red in Fig. 19) and successive bifurcations from which were appended Orders 2 through 5. Order 2 (orange in Fig. 20) are the branches of the Order 1 channels; and Order

3 channels are branches of the Order 2 (green in Fig. 20), and so on.

3.4.1 Migration of the Boundaries of the Andros Intertidal Channelized Zone Through

Time

Digitization of the coastline and inland Scytonema marsh line, which demarks the proximal and distal boundaries of the intertidal-channelized zone, respectively, reveals migrations that can be related to changes in sea level over the last 33 years (Fig. 21). Note that ‘0’ on the x-axes of Figure 21 B-through-E demarks the position of the coastlines and inland Scytonema marsh lines in 1985. Positive values on the x-axes of these plots represent the case that the boundaries of the intertidal channelized zone (see Fig. 15) migrate in an island-ward direction (i.e. retrogradation), relative to their position in 1985, whereas negative values represent seaward migrations (progradation). Starting in the north of

Andros Island, the Scytonema marsh line in the Triple Goose Creek focus area migrated up to 80 m island-ward between 1985 and 2012. Since the coastline of this focus area only

migrated island-ward by ~40 m, the breadth of the intertidal-channelized zone of Triple

Goose Creek increased in this time period. The three remaining focus areas - Blue Creek,

Middle Bight, and Cormorant Point – behave in accordance with one another and are different to the behavior observed in Triple Goose Creek. For these three areas, the width of the intertidal-channelized zones remained unchanged from 1985 through 2012. After

2012, however, the Scytonema marsh lines have migrated eastward, i.e. towards Andros

Island by 70 m, 220 m, and 190 m in the Blue Creek, Middle Bight, and Cormorant Point focus areas, respectively. In contrast the coastlines of these three focus areas remained unchanged in this time period. Consequently, this migration produces a meaningful broadening of the intertidal-channelized zone (and therefore narrowing of the supratidal inland Scytonema marsh zone which lies inboard of it). Over plotting the arrival of

Hurricanes which impacted Andros Island in the period of observation, which are

Hurricanes Andrew (1992), Michelle (2001), and Matthew (2016), which had categories 5,

4, and 5, respectively, supports the notion proposed by Rankey et al.(2004) that hurricanes do not exert meaningful sway over the evolution of the architecture of the Andros tidal flats. Comparing the movement of the boundaries of the intertidal-channelized zone with prevailing sea level (Fig. 21A), however, shows a conspicuous interrelationship. The pronounced expansion of the intertidal-channelized zone from 2011 onwards corresponds to a subtle but distinct increase in the rate of sea-level rise, and increase in overall sea-level amplitude which peaked in 2015. Note that the expansion of the intertidal channelized zone slowed at this time in the Blue Creek focus area, but continued unabated in those situated at Middle Bight and Cormorant Point.

Figure 21. (A) Shows the sea-level history in the region for the period 1985-2018 as quantified by the three most proximal research-grade tidal gauges to Andros Island (Key West, Virginia Key, and Settlement Point). With all situated within 300 km of Andros Island, data from these sites is considered representative of the long-term sea-level trend for Andros Island. (B-E) Chart the migration of the boundaries of the inter-tidal channelized zone in the four focus areas normalized to their position in 1985. Values <0 on the x-axes of these plots denote seaward migration of the channelized zone, whereas values >0 represent island-migration of the zone boundaries. Horizontal blue lines in (A) through (E) mark the arrival of major hurricanes that impacted Andros Island in the period of observation. Temporal changes in the patterning of Scytonema patches within the four focus areas was assessed on the basis of annual tallies of their quantity and size within the four focus areas (Fig. 22). As for the migration of the boundaries of the intertidal-channelized zone, the patch dynamics within the four areas cannot be related to hurricane impacts, but they do develop in-step with the assembled sea-level curves. The relationship to sea level is, however, more nuanced than for the migration of the boundaries of the channelized zone.

For instance, the major 2012-2015 episode of sea-level rise was proceeded by a lesser rise which peaked in 1999 (Fig. 22A). Although this first peak does not impact the breadth of the intertidal-channelized zone (i.e. Fig. 21), it does initiate a marked decrease in the size of the zone’s Scytonema patches, a trend consistent with the early stages of the fragmentation of this landscape. The second sea-level peak (2012 onwards) renews, indeed

accelerates, the decrease in area of the patches across all four focus areas, but also precipitates a radical increase in the number of patches (Fig. 22B through E). This trend can be interpreted to represent the wholescale fragmentation of the supratidal highs in the intertidal-channelized zone (for which the presence of Scytonema is a proxy), a change induced by the fact that sea level has risen beyond a critical threshold for the stability of this landscape to be maintained.

Figure 22. (A) Shows the sea level history in the region for the period 1985-2018. The inter-tidal channelized zone of the Andros tidal flats is characterized by Scytonema patches, the distribution of which are tracked over the period of observation in terms of their area and number (B) through (E). The timing of hurricanes, over-plotted with horizontal blue lines, appear to have little impact on patch patterning, whereas increases in sea-level, first in 2001 and again in 2011, appear to exert meaningful control, leading to fragmentation of the Scytonema landscape (text for details). 3.4.2 Changes Through Time in the Architecture of the Andros Tidal Channels at Triple

Goose Creek

Analysis of the evolution of the tidal creeks in the vicinity of Triple Goose Creek reveals substantial changes occurred between 1943 and 2018. As captured by Figure 23C, in all focus areas save for that situated in the southern reaches of Triple Goose Creek, the length of the channels created between 1943 and 2018 greatly exceeds those lost in the same period due to channel abandonment. Hence, the channel network in these three areas

is becoming more extensive with time (and with rising sea level). Partitioning the data into five orders of bifurcation reveals that the increase in the length of channels is concentrated in bifurcation Orders 3 and 4. No changes are observed for Order 1. The southernmost focus area is distinct in that there is a >1.5 km reduction in the length of its third order channels, which exceeds the sum of the lengths of the increased extension of the third through fifth order channels. However, if the exceptional third order behavior is excluded, the length of fourth and fifth order channels created are longer than those abandoned. If the behavior of the tidal channels is plotted against the breadth of the intertidal-channelized zone (Fig. 23D) and direction that the coastline of the four focus areas face (Fig. 23E), it can be seen that the two focus areas whose channels are the most dynamic through time

(i.e. the pair situated in the middle of Triple Goose Creek) are also the areas where the channelized zone is at its broadest, and those with west-facing coastlines therefore normal to the passage of winter cold fronts. These observations suggest that increased channel dynamics might be anticipated in areas where the intertidal-channelized zone is best developed, and if the zone broadens through time (as suggested by Fig. 21), that channel dynamics will increase in step with the sea level transgression.

3.5.1 Scytonema-Dominated Facies Respond Swiftly to Changing Sea Level

Scytonema is a large dark filamented (individual filaments 10 to 30 μm) cyanobacteria, primarily a freshwater lover, that grows in stubbly upright tufts up to several millimeters in length (Black, 1933; Monty, 1967; Hardie, 1977). The near surface sediment below the

Scytonema mat is densely layered from millimeter-laminae to thin beds up to 10 cm thick

(Fig. 23B) (Hardie, 1977). Tracking the change in the architecture of Scytonema mats through time serves as a sensitive proxy for even minor sea-level oscillations by virtue of the fact that this genus of multicellular filamentous cyanobacteria is biologically bound to

the high-elevation interface which separates the intertidal-channelized and supratidal zones of the Andros tidal flat complex (Shinn et al., 1965; Hardie, 1977) Such high-elevation areas allow these tidal flat portions to be relatively free from the day-to-day tidal fluctuations (Hardie, 1977). The organism creates vast black mats that extend 4-8 km eastward abutting the Andros pineland perched atop the exposed Pleistocene limestones of the island itself (Fig. 23A) (Shinn et al., 1965; Hardie, 1977). Although Scytonema patches also develop in the levee-pond complex in the intertidal-channelized zone, they dominate the island-ward portion of the intertidal-channelized zone and are fully developed within the supratidal zone. With the help of Google Earth Timelapse, the annual differences in the changes of Scytonema growth patterns can be detected using GIS tools.

After deducting the differences caused by the winter and summer seasons, the rate of change was calculated between the two facies: Scytonema-dominated facies including

Scytonema patches and the inland Scytonema marsh line, and the non-Scytonema- dominated facies of the coastline and remaining parts of the channelized zone. Their contrasting rates of change through time could then be related to external forcings such as sea level. With sea level rise, incoming seawater flows through the tidal channels and into the ponds, thereby increasing the salinity of any ponded water (Hardie, 1977). When pond salinity exceeds one part per thousand, the colonies of sensitive Scytonema marsh would decay in size and fragmentize (Black, 1933). Hardie (1977) showed that Scytonema can tolerate no more than 4-5 days of continuous submergence in seawater, and any colonization even a centimeter or two below the critical surface elevation, where the submergence time abruptly doubles between the mean high water neaps (MHWN) and mean tide level (MTL), will be exposed to at least one lethal dose (more than 9 days) of

seawater each year (see his Fig. 93). A ~10 cm sea level rise is sufficient to affect the intertidal-channelized zone and lead to the whole scale change of the Scytonema-dominated facies. Figs. 21B-E depicts the largest island-ward coastline migration of 40 m

(accomplished at a rate of ~1.2 m/yr.) at the Triple Goose Creek site, as compared to other three study areas; while the Scytonema marsh line migrated island-ward in the Middle

Bight with the largest distance of 220 m (~7 m/yr.). The range of the sea level rise (~10 cm) in this study is reassuringly comparable to that (~8 cm) observed by Rankey and

Morgan (2002). Though the island-ward migration (erosion) of the coastline is anticipated as a consequence of sea level rise (Pethick, 2001), it is mostly due to the mixed and complex physical-induced process that depends on the offshore hydrodynamic sub- environments (Kench and Mann, 2017; Duvat, 2019).

Additionally, Black (1933), Gebelein (1969), and Hardie (1977) emphasized a vital role played by an additional smooth, tough and rubbery mat called Schizothrix (Fig. 24C).

This pink-gray cyanobacteria with filaments <5μm covers seaward portions of the intertidal channelized zone including beach terraces and ridges. This type of mat is an ingrained pervasive sediment-binding filament complex, from which the sub-environments develop laminations ranging from 0.1-2 mm (see Hardie 1977 in Fig. 34). According to the observation of Black (1933), Gebelein (1969), and Hardie (1977), Schizothrix develops to form soft gelatinous mats within which tiny particles ranging from mud to grain size are trapped and bound extremely tight that it protects the underlying sediment from erosion during hurricane floodings or rain wash. Therefore, the cohesive nature of sediments along the coastline partly explains its static character, as captured in Figure 21. Schizothrix, which also occurs on the channel levees, may locally retard erosion and migration of channels.

However, Schizothrix patches are difficult to monitor through the remote sensing data as the patches of Schizothrix along the beach terrace and levee crests are narrow (10-30 m).

Figure 24. (A) A view of surface of Scytonema marsh. (B) Shows the surface of sediment below the Scytonema marsh, which is densely layered from a millimeter-laminae to thin beds. (C) Shows the light color Schizothrix mats in the levees. Located in the north of the Tropic of Cancer, Andros Island is exceptionally wet and rainy (Hardie, 1977). Though not being detailed recorded in the previous studies, the volume of rainfall does significantly affect the evolution of morphology (such as flooding pattern) in the shallow tidal flat. Hardie (1977) pointed out the mat-morphology changes within any Scytonema-dominated facies correlate with the wetting. Residual seawater in land would be diluted by rainwater to a great extent. As, emphasized by Hardie (1977), the volume of rainfall directly determines the mat thickness and continuity. During the rainy period, three-quarters of the rain's annualized volume fall into the intertidal-channelized zone and keep ponds far from draining (Hardie, 1977). Groundwater headed to the eastern pinelands humidifies the inland Scytonema mats with fresh water in summer (Hardie, 1977).

Annualized rainfall data were collected in the Andros Town (Fig. 24A). Fig. 25B captures the rainfall trend provided by World Weather Online (data provided by

WorldWeatherOnline.com; used with permission) from 2009-2018. However, such annualized data is two times lower than the data Hardie collected from U.S. Naval

Oceanographic Office Publ. H. O. 128 (see his Table 1). Hardie (1977) recorded an annual

rainfall average of about 129 cm a year and can be as high as 232 cm or as low as 65 cm.

Thus, Andros Island was comparatively arid in the past decade, which might accelerate the fragmentation of Scytonema mats. Therefore, long-term rainfall data needs to be combined with a sea-level change rate to evaluate the change of morphology on the tidal flat, which will be the new topic for future scholars.

Figure 25. (A) Shows the location of the Andros Town where the rainfall data were collected. (B) Plots the annualized rainfall amount from 2009-2018. 3.5.2 Transgression in the Rock Record

This study constitutes one of few that have quantified the dynamic changes of boundaries of tidal flats (the coastline and an internal boundary separating the intertidal from supratidal portions) at timescales relevant to the projections of the tidal-flat architecture in the rock record. The results of this study show that the changes in the positions of these two boundaries are the result of sea level rise, especially at the key internal boundary. As the intertidal-channelized zone widens by extending further into the supratidal area, subtle changes leave evidence that would potentially allow sedimentologists to recognize the earliest transgression in the rock record.

Combined with the dynamics of the boundaries of the intertidal-channelized zone

(Figs. 21, 22, and 23) and fluctuations of sea-level change (Fig. 17), three hypothetical cores were designed to simulate the earliest transgression in the rock record (Fig. 25A-C).

The first hypothetical core is placed at the most seaward location of the coastline. The results in Figure 21 show that the coastline remained relatively static after the establishment of the tidal flats even when sea level continued to rise. Therefore, a change (such as the rock type or sediment structure) is unlikely to be recorded in this first location. At the coast

(location of the first core), however, erosion of the beach ridge and of the underlying mudstone has occurred, and in this location, the presence of intraclasts and minor transgression of subtidal facies should therefore be anticipated. Hence, in this hypothetical core, laminations which represent beach ridge deposits will decrease up-core, accompanied by an increase in the presence of more bioturbated mud. More noticeable signs of the transgression are found in the second and third hypothetical cores, which are located at the center and landward portions of the channelized zone, respectively (Fig. 25A-B). Two types of evidence for a transgression should be anticipated at the position of the second hypothetical core: (i) channel deposits (e.g. intraclasts) might be produced in the new layer of sediments corresponding to a lengthening of the channels, and, (ii) a decrease of the prevalence of the laminated Scytonema marsh up the core section due to the fragmentation and reduction of the Scytonema patches. Considering the hundreds of meters of migration of the landward boundary of the channelized zone (as demonstrated in Fig. 20), the third hypothetical core would likely represent the site of greatest change caused by the transgression, wherein the thinly bedded and laminated Scytonema marsh would change

into a facies more similar to the intertidal-channelized zone, which is characterized by bioturbated mudstone, patches of Scytonema, and localized channel deposits.

These hypothetical cored records are partially reinforced by a core sample collected in one of the abandoned channels in the Triple Goose Creek (Fig. 27) in April 2019.

Gastropod shells in Fig. 27G are considered as the signature of the abandoned channel.

Other actual core descriptions of the layering in the Andros tidal flat (e.g. Black, 1933;

Shinn et al., 1969) as well as other modern carbonate tidal flats (e.g. Illing et al., 1965;

Davies, 1970) also support the hypothetical cored records. The laminations in the records represent a response to recurrent or rhythmic change in the sub-environment of alternate flooding and draining (Black, 1933). Illing et al. (1965) noted the exclusive development of lamination at the intertidal cyanobacterial mats in the Qatar area of the Persian/Arabian

Gulf. They reported that these ‘stromatolitic algal laminations’ could be used to recognize the buried tidal flat environment upon the supratidal sabkha. Meanwhile, Shinn et al.

(1969) observed that the cores taken in the adjacent marine, bar, channel, and pond were lack of layering, whereas the laminations in the levee are well developed and accumulated patchily (see their Figs. 3, 6, 9, and 19, respectively). Kendall and Skipwith (1968), and

Davies (1970a) also made the same general observation. Davies (1970a) specifically described two types of layering that were associated with microbial mats in detail: (i) High sediment volume with low organic content in the outer intertidal zone; (ii) Moderate sedimentation with high organic content in the blocked channels. Logan et al. (1974) emphasized the importance of mat types in determining the fabric of the sedimentation.

Hardie (1977) summarized the mechanisms of layering in the Triple Goose Creek tidal flat sediments and concluded that the mean tide level (MTL) could be served as a distinction

for sediment distribution. Below the MTL sediments are essentially unlayered and homogenized by burrowing organisms. Above the MTL layering is well preserved as (1) thin (1-2-mm) laminations, (2) thin beds and thick laminations restricted to the mat, and

(3) cross-bedded skeletal sands (on beach ridges). Hardie (1977) stressed these distinctive types of layering matches with their sub-environments. By examining geological records from extensive sediment coring, Maloof and Grotzinger (2012) generated the evolution process of Triple Goose Creek. Only the well-laminated cyanobacteria mats could represent distinctive primary depositional processes. For example, Schizothrix laminations near channels are defined as the first obvious indicator of an upward-shallowing trend.

However, it is worth noting that the small changes in the water depth control massive-scale changes in the sub-environment, though the secondary depositional processes might obscure that. The cumulation of various hypotheses and descriptions of layering in the tidal flats might be compared and contrasted with change in water depth to enable a more accurate evolution process of ancient carbonates.

Figure 26. (A-B) shows the locations of the three hypothetical cores in both vertical and horizontal respects in the intertidal-channelized zone: Core 1 locates in the proximal location of the coastline; Core 2 in the middle point of the intertidal-channelized zone; Core 3 locates in the distal portion of the intertidal-channelized zone. Based on the results of our time-separated remote sensing, (C) shows the idealized rock record in the three- hypothetical cores prior to the initiation of the earliest transgression, during the transgression, and immediately after it.

Chapter 4: Conclusion

This thesis has considered the behavior of two sets of islands over a period of 50+ years. First, the islands of the Chagos Archipelago, which are situated on the platform margin and therefore exposed to the prevailing hydrodynamic conditions of the open Indian

Ocean. The second case study is Andros Island, in the Bahamas, which, in stark contrast to the Chagos islands, is situated in the interior of its host platform (Great Bahama Bank) and therefore in a more quiescent hydrodynamic setting. Against this backdrop, it might logically be anticipated that the coastlines of the Chagos islands would be more dynamic than those of Andros Island. In fact, this is not the case – both case studies emphasize stability to the islands’ coastlines through time, even in the face of considerable (10+ cm) sea-level rise during the periods of observation. This is not to say that the coastlines are static. This is particularly the case for the coastlines of the islands of Peros Banhos Atoll, which are aligned to face the prevailing hydrodynamic energy. These coastlines tend to retreat through time, whereas those aligned away from prevailing energy, expand. In aggregate, this behavior means that, through time, the Peros Banhos islands migrate across the atoll rim which hosts them, down-dip, following the prevailing gradient of hydrodynamic energy. The long-term trends will be easily predicted through amplifying trends of short-term shoreline mobility. Human intervention and specific morphological elements such as embayments in Diego Garcia are seen to amplify coastline dynamics.

In an attempt to relate the behavior of atoll-islands through time to global patterns of sea-level rise, data from the Chagos was merged with a global database capturing the behavior of atoll islands developed by Duvat (2019) (Fig. 14). Surprisingly, this

comparison shows that atoll islands situated in areas with high rates of sea-level rise are no more dynamic than those situated in areas where the rate of rise is lower. The behavior of the supratidal beaches of Earth’s atoll islands in the face of rising sea level contrasts drastically with the behavior of the inter- and supratidal deposits of Andros Island. Three caveats are important here, however. First, in the Chagos, the behavior through time of the coarse grainstone deposits of the islands’ beaches was tracked. The deposits considered in

Andros Island are those which comprise the inter- and supratidal mud flats, which are composed of cohesive aragonitic mudstone. The second caveat is that the Chagos data come from the supratidal (beach) deposits only, whereas in Andros Island both the behavior of the beach and the expansive inter- and supratidal mud flats which lay inboard of it were tracked. Third, the Chagos Archipelago lies too close to the equator to experience cyclones.

This contrasts with Andros Island, situated north of the Tropic of Cancer, in the hurricane belt, which was impacted by 3 storms with magnitudes exceeding Category 4 over the seventy-year period of observation.

The Chagos atoll islands, which are situated on the platform margin, and Andros

Island (situated in the platform interior) are consistent in that their coastlines do not respond predictably to rising sea level over the last five decades. Instead, prevailing hydrodynamic energy better explains coastline behavior in Peros Banhos. Moving to the Andros tidal flats, however, the data suggests that this environment responds swiftly and dramatically to rising sea level or even the volume of rainfall. The response of the tidal flats to the change of external environment is threefold. First, in the intertidal-channelized zone, which lies directly inboard from the beach, any supratidal deposits that are characterized by their dense-dark Scytonema mats fragment in the face of rising sea level. This fragmentation is

marked by a decrease in the aerial extent of the Scytonema mats and an increase in their degree of patchiness. Second, and in the same zone, the tidal channels which bisect the

(aptly) named intertidal-channelized zone, avulse and lengthen in the face of rising sea level. This is not to say that some channels become abandoned due to siltation, which does occur, but more channels are created via avulsion, than lost to siltation, during the period of observation. The third response of the Andros tidal flats to a 10 cm sea-level rise is that the island-ward boundary of the intertidal-channelized zone, prograde up to 200 m over the more distal supratidal-inland Scytonema mat zone.

With an eye to the ancient, the dynamics of the Chagos islands over the last 50 years are unlikely to be captured in the rock record. Their behavior would lead to a grainy (beach) cap atop the reef-flat framestone matrix, for the case of a hypothetical core sunk in proximity to an island and down-dip to the prevailing energy gradient. In short, the earliest part of a sea-level transgression is unlikely to be captured by an atoll-island sequence. This stands in stark contrast to the platform-interior tidal flats of Andros Island. While the position of the coastline is static in the face of a 10 cm sea-level rise, the fragmentation of the supratidal cyanobacterial mats inboard from the coast would mean that the earliest transgression is demarked by a radical decrease in the prevalence of stromatolitic fabrics in this zone. Farther inboard, meanwhile, and due to its modest topographic relief, the supratidal zone of Andros will be capped by the amorphous mud of the intertidal zone, and, possibly also ornamented by channel-lag deposits created by the extension and avulsion of the tidal-creek network. In aggregate, the direct response of the tidal flats to even 10 cm of sea-level rise considerably elevates the likelihood that the very earliest transgression in an analogous environment would have a signature in the rock record. However, it should

always be noted that sea-level change alone cannot explain all the lateral complexity and regional variabilities occurred in the tidal flats.

References

Albert, S., Leon, J.X., Grinham, A.R., Church, J.A., Gibbes, B.R. and Woodroffe, C.D., 2016. Interactions between sea-level rise and wave exposure on reef island dynamics in the Solomon Islands. Environmental Research Letters, 11(5), p.054011. Andréfouët, S., Guillaume, M.M.M., Delval, A., Rasoamanendrika, F.M.A., Blanchot, J., Bruggemann, J.H., 2013. Fifty years of changes in reef flat habitats of the Grand Récif of Toliara (SW Madagascar) and the impact of gleaning. Coral Reefs 32, 757–768. Andréfouët, S., Muller-Karger, F.E., Robinson, J.A., Kranenburg, C.J., Torres-Pulliza, D., Spraggins, S.A. and Murch, B., 2006, June. Global assessment of modern coral reef extent and diversity for regional science and management applications: a view from space. In Proceedings of the 10th International Coral Reef Symposium (Vol. 2, pp. 1732-1745). Japanese Coral Reef Society Okinawa, Japan. Aslam, M. and Kench, P.S., 2017. Reef island dynamics and mechanisms of change in Huvadhoo Atoll, Republic of Maldives, Indian Ocean. Anthropocene, 18, pp.57-68. Baart, F., Van Gelder, P.H., De Ronde, J., Van Koningsveld, M. and Wouters, B., 2012. The effect of the 18.6-year lunar nodal cycle on regional sea-level rise estimates. Journal of Coastal Research, 28(2), pp.511-516. Bailey, R.T., 2015. Quantifying transient post‐overwash aquifer recovery for atoll islands in the Western Pacific. Hydrological Processes, 29(20), pp.4470-4482. Barkey, B.L. and Bailey, R.T., 2017. Estimating the impact of drought on groundwater resources of the Marshall Islands. Water, 9(1), p.41. Beetham, E., Kench, P.S. and Popinet, S., 2017. Future reef growth can mitigate physical impacts of sea‐level rise on atoll islands. Earth's Future, 5(10), pp.1002-1014. Beven, Jack; National Hurricane Center (January 23, 2002). Hurricane Michelle (PDF) (Tropical Cyclone Report). Miami, Florida: United States National Oceanic and Atmospheric Administration's National Weather Service. Retrieved 20 June 2013. Biribo, N. and Woodroffe, C.D., 2013. Historical area and shoreline change of reef islands around Tarawa Atoll, Kiribati. Sustainability Science, 8(3), pp.345-362. Black, M., 1932. VI. The algal sediments of Andros Island, Bahamas. Philosophical Transactions of the Royal Society of London. Series B, Containing Papers of a Biological Character, 222(483-493), pp.165-192. Boberg, J., 2009. Litter decomposing fungi in boreal forests (Vol. 2009, No. 75). Chollett, I. and Mumby, P.J., 2012. Predicting the distribution of Montastraea reefs using wave exposure. Coral Reefs, 31(2), pp.493-503.

Church, J.A., White, N.J. and Hunter, J.R., 2006. Sea-level rise at tropical Pacific and Indian Ocean islands. Global and Planetary Change, 53(3), pp.155-168. Coleman, D.C., Callaham, M.A. and Crossley Jr, D.A., 2017. Fundamentals of soil ecology. Academic Press. Collen, J.D., Garton, D.W., Gardner, J.P.A., 2009. Shoreline changes and sediment redistribution at Palmyra atoll (equatorial Pacific Ocean): 1874–present. J. Coast. Res. 25, 711–722. Davies, G.R., 1970. Algal-laminated sediments, Gladstone Embayment, Shark Bay, Western Australia. Dickinson, W.R., 1999. Holocene sea-level record on Funafuti and potential impact of global warming on central Pacific atolls. Quaternary Research, 51(2), pp.124-132. Droxler AW, Jorry SJ. The Origin of Modern Atolls: Challenging Darwin's Deeply Ingrained Theory. Annual Review of Marine Science. 2020 Sep. DOI: 10.1146/annurev- marine-122414-034137. Dunne, R.P., Barbosa, S.M. and Woodworth, P.L., 2012. Contemporary sea level in the Chagos Archipelago, central Indian Ocean. Global and Planetary Change, 82, pp.25-37. Duvat, V.K. and Pillet, V., 2017. Shoreline changes in reef islands of the Central Pacific: Takapoto Atoll, Northern Tuamotu, French Polynesia. Geomorphology, 282, pp.96-118. Duvat, V.K., 2019. A global assessment of atoll island planform changes over the past decades. Wiley Interdisciplinary Reviews: Climate Change, 10(1), p.e557. Duvat, V.K., 2020. Human-driven atoll island expansion in the Maldives. Anthropocene, 32, p.100265. Duvat, V.K., Salvat, B. and Salmon, C., 2017. Drivers of shoreline change in atoll reef islands of the Tuamotu Archipelago, French Polynesia. Global and planetary change, 158, pp.134-154. Eisenhauer, A., Heiss, G.A., Sheppard, C.R.C. and Dullo, W.C., 1999. Reef and island formation and Late Holocene sea-level changes in the Chagos islands. Ecology of the Chagos Archipelago, pp.21-31. Ekebom, J., Laihonen, P. and Suominen, T., 2003. A GIS-based step-wise procedure for assessing physical exposure in fragmented archipelagos. Estuarine, Coastal and Shelf Science, 57(5-6), pp.887-898. Fagherazzi, S., 2008. Self-organization of tidal deltas. Proceedings of the National Academy of Sciences, 105(48), pp.18692-18695. Fagherazzi, S., Carniello, L., D'Alpaos, L. and Defina, A., 2006. Critical bifurcation of shallow microtidal landforms in tidal flats and salt marshes. Proceedings of the National Academy of Sciences, 103(22), pp.8337-8341.

Falkland, T. and White, I., 2020. Freshwater Availability Under Climate Change. In Climate Change and Impacts in the Pacific (pp. 403-448). Springer, Cham. Flowers, T.J. and Colmer, T.D., 2015. Plant Salt Tolerance: Adaptations in Halophytes. Annals of Botany, 115(3), pp.327-331. Ford, M., 2012. Shoreline changes on an urban atoll in the central Pacific Ocean: Majuro Atoll, Marshall Islands. Journal of Coastal Research, 28(1), pp.11-22. Ford, M., 2013. Shoreline changes interpreted from multi-temporal aerial photographs and high-resolution satellite images: Wotje Atoll, Marshall Islands. Remote Sensing of Environment, 135, pp.130-140. Ford, M.R. and Kench, P.S., 2015. Multi-decadal shoreline changes in response to sea level rise in the Marshall Islands. Anthropocene, 11, pp.14-24. Ford, M.R. and Kench, P.S., 2016. Spatiotemporal variability of typhoon impacts and relaxation intervals on Jaluit Atoll, Marshall Islands. Geology, 44(2), pp.159-162. Ford, M.R., Kench, P.S., 2014. Formation and adjustment of typhoon-impacted reef islands interpreted from remote imagery: Nadikdik atoll, Marshall Islands. Geomorphology 214, pp. 216–222. Gebelein, C.D., 1969. Distribution, morphology, and accretion rate of recent subtidal algal stromatolites, Bermuda. Journal of Sedimentary Research, 39(1), pp.49-69. Ginsburg, R.N. and Hardie, L.A., 1975. Tidal and storm deposits, northwestern Andros Island, Bahamas. In Tidal deposits (pp. 201-208). Springer, Berlin, Heidelberg. Ginsburg, R.N., 1974, Introduction to comparative sedimentology of carbonates: American Association of Petroleum Geologists, Bulletin, v. 58, pp. 781–786. Goldberg, J.N., MacFarlane, A.J., Newman, E.T., Rohrlich, F. and Sudarshan, E.G., 1967. Spin‐s Spherical Harmonics and ð. Journal of Mathematical Physics, 8(11), pp.2155-2161. Graham, N.A., Wilson, S.K., Carr, P., Hoey, A.S., Jennings, S. and MacNeil, M.A., 2018. Seabirds enhance coral reef productivity and functioning in the absence of invasive rats. Nature, 559(7713), pp.250-253. Hamylton, S. and East, H., 2012. A Geospatial Appraisal of Ecological and Geomorphic Change on Diego Garcia Atoll, Chagos Islands (British Indian Ocean Territory). Remote Sensing, 4(11), pp.3444-3461. Hardie, L.A. ed., 1977. Sedimentation on the modern carbonate tidal flats of northwest Andros Island, Bahamas (Vol. 22). Johns Hopkins University Press. Harris, P.M., Purkis, S.J. and Ellis, J., 2011. Analyzing spatial patterns in modern carbonate sand bodies from Great Bahama Bank. Journal of Sedimentary Research, 81(3), pp.185- 206.

Harris, P.M., Purkis, S.J., Ellis, J., Swart, P.K. and Reijmer, J.J., 2015. Mapping bathymetry and depositional facies on Great Bahama Bank. Sedimentology, 62(2), pp.566- 589. Hazard, C.S., Ritter, S.M., McBride, J.H., Tingey, D.G. and Kench II, R.W., 2017. Ground- penetrating-radar characterization and porosity evolution of an Upper Pleistocene oolite- capped depositional cycle, Red Bays, northwest Andros Island, Great Bahama Bank. Journal of Sedimentary Research, 87(5), pp.523-545. Hernández-Cruz LR, Purkis SJ, Riegl BM (2006) Documenting decadal spatial changes in seagrass and Acropora palmata cover by aerial photography analysis in Vieques, Puerto Rico: 1937–2000. Bulletin of Marine Science 79(2), pp. 401–414 Holdaway, A. and Ford, M., 2019. Resolution and scale controls on the accuracy of atoll island shorelines interpreted from satellite imagery. Applied Geomatics, 11(4), pp.339-352. Hubbard, D., Gischler, E., Davies, P., Montaggioni, L., Camoin, G., Dullo, W.-C., Storlazzi, C., Field, M., Fletcher, C., Grossman, E., Sheppard, C., Lescinsky, H., Fenner, D., McManus, J., Scheffers, S., 2014. Island outlook: warm and swampy. Science 19, pp. 1461. Illing, L.V. and Wells, A.J., 1964. Penecontemporary dolomite in the Persian Gulf. AAPG Bulletin, 48(4), pp.532-533. Jahnert, R.J. and Collins, L.B., 2013. Controls on microbial activity and tidal flat evolution in Shark Bay, Western Australia. Sedimentology, 60(4), pp.1071-1099. Kahma, K.K. and Calkoen, C.J., 1992. Reconciling discrepancies in the observed growth of wind-generated waves. Journal of Physical Oceanography, 22(12), pp.1389-1405. Kaluwin, C., Smith, A., 1997. Coastal vulnerability and integrated coastal zone management in the South Pacific Island region. J. Coast. Res. 24, pp. 95–106. Kench, P.S. and Brander, R.W., 2006a. Response of reef island shorelines to seasonal climate oscillations: South Maalhosmadulu atoll, Maldives. Journal of Geophysical Research: Earth Surface, 111(F1). Kench, P.S. and Mann, T., 2017. Reef island evolution and dynamics: insights from the Indian and Pacific Oceans and perspectives for the Spermonde Archipelago. Frontiers in Marine Science, 4, p.145. Kench, P.S., Brander, R.W., 2006b. Wave processes on coral reef flats: implications for reef geomorphology using Australian case studies. J. Coast. Res. 22, pp. 209–223. Kench, P.S., Ford, M.R. and Owen, S.D., 2018. Patterns of Island Change and Persistence Offer Alternate Adaptation Pathways for Atoll Nations. Nature communications, 9(1), pp. 1-7. Kench, P.S., McLean, R.F., Nichol, S.N., 2005. Newmodel of reef-island evolution: Maldives, Indian Ocean. Geology 33, pp. 145–148.

Kench, P.S., Thompson, D., Ford, M.R., Ogawa, H., McLean, R.F., 2015. Coral islands defy sea-level rise over the past century: records from a Central Pacific Atoll. Geology 43, pp. 515–518. Kendall, C.G.S.C. and d'E. Skipwith B, P.A., 1968. Recent algal mats of a Persian Gulf lagoon. Journal of Sedimentary Research, 38(4), pp. 1040-1058. Laporte, L.F., 1967. Carbonate deposition near mean sea-level and resultant facies mosaic: Manlius Formation (Lower Devonian) of New York State. AAPG Bulletin, 51(1), pp.73- 101. Le Cozannet, G., Garcin, M., Petitjean, L., Cazenave, A., Becker, M., Meyssignac, B., Walker, P., Devilliers, C., Le Brun, O., Lecacheux, S. and Baills, A., 2013. Exploring the relation between sea level rise and shoreline erosion using sea level reconstructions: an example in French Polynesia. Journal of Coastal Research, 65(sp2), pp. 2137-2142. Logan, B.W., Hoffman, P. and Gebelein, C.D., 1974. Algal mats, cryptalgal fabrics, and structures, Hamelin Pool, Western Australia. American Association of Petroleum Geologists, Memoir 22: 140–194. Maloof, A.C. and Grotzinger, J.P., 2012. The Holocene shallowing‐upward parasequence of north‐west Andros Island, Bahamas. Sedimentology, 59(4), pp.1375-1407. Masselink, G. and Pattiaratchi, C.B., 2001. Seasonal Changes in Beach Morphology along the Sheltered Coastline of Perth, Western Australia. Marine geology, 172(3-4), pp.243-263. McGowan, A., Broderick, A.C. and Godley, B.J., 2008. Seabird populations of the Chagos Archipelago: an evaluation of Important Bird Area sites. Oryx, 42(3), pp. 424-429. McLean, R. and Kench, P., 2015. Destruction or persistence of coral atoll islands in the face of 20th and 21st century sea‐level rise? Wiley Interdisciplinary Reviews: Climate Change, 6(5), pp. 445-463. Monty, C.L., 1972. Recent algal stromatolitic deposits, Andros Island, Bahamas. Preliminary report. Geologische Rundschau, 61(2), pp. 742-783. Morgan, K.M. and Kench, P.S., 2016. Parrotfish erosion underpins reef growth, sand talus development and island building in the Maldives. Sedimentary Geology, 341, pp. 50-57. Newnham, T.J., Browne, N.K., Bumbak, J., Loudon, L., Wellington, H., Shedrawi, G., Hacker, J. and O’Leary, M., 2020. Long-term (70-year) monitoring of reef structure through high-resolution multidecadal aerial imagery. Coral Reefs, pp.1-12. Nicholls, R.J. and Cazenave, A., 2010. Sea-level Rise and Its Impact on Coastal Zones. science, 328(5985), pp. 1517-1520. Perry, C.T. and Morgan, K.M., 2017. Bleaching drives collapse in reef carbonate budgets and reef growth potential on southern Maldives reefs. Scientific Reports, 7, p. 40581.

Perry, C.T., Alvarez-Filip, L., Graham, N.A., Mumby, P.J., Wilson, S.K., Kench, P.S., Manzello, D.P., Morgan, K.M., Slangen, A.B., Thomson, D.P. and Januchowski-Hartley, F., 2018. Loss of coral reef growth capacity to track future increases in sea level. Nature, 558(7710), pp. 396-400. Perry, C.T., Kench, P.S., O'Leary, M.J., Morgan, K.M., Januchowski-Hartley, F., 2015. Linking reef ecology to island building: parrotfish identified as major producers of island- building sediment in the Maldives. Geology 43, pp. 503–506. Perry, C.T., Kench, P.S., Smithers, S.G., Riegl, B., Yamano, H. and O'Leary, M.J., 2011. Implications of reef ecosystem change for the stability and maintenance of coral reef islands. Global Change Biology, 17(12), pp. 3679-3696. Perry, C.T., Kench, P.S., Smithers, S.G., Yamano, H., O’Leary, M. and Gulliver, P., 2013. Time scales and modes of reef lagoon infilling in the Maldives and controls on the onset of reef island formation. Geology, 41(10), pp. 1111-1114. Pethick, J., 2001. Coastal management and sea-level rise. Catena, 42(2-4), pp.307-322. Purkis SJ, Rowlands G, Kerr JM (2015) Unravelling the influence of water depth and wave energy on the facies diversity of shelf carbonates. Sedimentology. 62:541-565 Purkis, S.J. and Harris, P.M., 2016. The extent and patterns of sediment filling of accommodation space on Great Bahama Bank. Journal of Sedimentary Research, 86(4), pp.294-310. Purkis, S.J. and Klemas, V., 2011. Remote sensing and global environmental change. John Wiley & Sons. Purkis, S.J., Gardiner, R., Johnston, M.W. and Sheppard, C.R., 2016. A half-century of coastline change in Diego Garcia–The largest atoll island in the Chagos. Geomorphology, 261, pp. 282-298. Purkis, S.J., Graham, N.A.J. and Riegl, B.M., 2008. Predictability of Reef Fish Diversity and Abundance Using Remote Sensing Data in Diego Garcia (Chagos Archipelago). Coral reefs, 27(1), pp.167-178. Purkis, S.J., Kohler, K.E., Riegl, B.M. and Rohmann, S.O., 2007. The statistics of natural shapes in modern coral reef landscapes. The Journal of Geology, 115(5), pp. 493-508. Quataert, E., Storlazzi, C., Van Rooijen, A., Cheriton, O. and Van Dongeren, A., 2015. The influence of coral reefs and climate change on wave‐driven flooding of tropical coastlines. Geophysical Research Letters, 42(15), pp. 6407-6415. Rankey, E.C. and Morgan, J., 2002. Quantified rates of geomorphic change on a modern carbonate tidal flat, Bahamas. Geology, 30(7), pp.583-586.

Rankey, E.C., 2002. Spatial patterns of sediment accumulation on a Holocene carbonate tidal flat, northwest Andros Island, Bahamas. Journal of Sedimentary Research, 72(5), pp.591-601. Rankey, E.C., 2011. Nature and stability of atoll island shorelines: Gilbert Island chain, Kiribati, equatorial Pacific. Sedimentology 58, pp. 1831–1859. Rankey, E.C., Enos, P., Steffen, K. and Druke, D., 2004. Lack of impact of hurricane Michelle on tidal flats, Andros Island, Bahamas: integrated remote sensing and field observations. Journal of Sedimentary Research, 74(5), pp.654-661. Rappaport, E., 1993. National Hurricane Center (Preliminary Report), Miami, FL. Richmond, R.H., 1993. Coral reefs: present problems and future concerns resulting from anthropogenic disturbance. American Zoologist, 33(6), pp. 524-536. Roehl, P.O., 1967. Stony Mountain (Ordovician) and Interlake (Silurian) facies analogs of Recent low-energy marine and subaerial carbonates, Bahamas. AAPG Bulletin, 51(10), pp.1979-2032. Roy, P. and Connell, J., 1991. Climatic change and the future of atoll states. Journal of Coastal Research, pp. 1057-1075. Schlager, W., 1993. Accommodation and Supply—a Dual Control on Stratigraphic Sequences. Sedimentary geology, 86(1-2), pp. 111-136. Sheppard C, Sheppard A, Mogg A, Bayley D, Dempsey A, Roche R, Turner J, Purkis SJ (2017) Coral bleaching and mortality in the Chagos Archipelago to 2017. Atoll Research Bulletin 613, pp.1:26 Sheppard, C., Dixon, D.J., Gourlay, M., Sheppard, A. and Payet, R., 2005. Coral mortality increases wave energy reaching shores protected by reef flats: examples from the Seychelles. Estuarine, Coastal and Shelf Science, 64(2-3), pp. 223-234. Sheppard, C., Sheppard, A. and Fenner, D., 2020. Coral mass mortalities in the Chagos Archipelago over 40 years: Regional species and assemblage extinctions and indications of positive feedbacks. Marine Pollution Bulletin, 154, p.111075. Sheppard, C.R., 2016. Changes to the natural history of islands in the Chagos atolls, central Indian Ocean, during human settlement (1780–1969), and prospects for restoration. Atoll Research Bulletin, 29, pp. 612. Sheppard, C.R., Ateweberhan, M., Bowen, B.W., Carr, P., Chen, C.A., Clubbe, C., Craig, M.T., Ebinghaus, R., Eble, J., Fitzsimmons, N. and Gaither, M.R., 2012. Reefs and islands of the Chagos Archipelago, Indian Ocean: why it is the world's largest no‐take marine protected area. Aquatic Conservation: marine and freshwater ecosystems, 22(2), pp. 232- 261.

Shinn, E.A., Ginsburg, R.N. and Lloyd, R.M., 1965. Recent supratidal dolomite from Andros Island Bahamas: in L. C. Pary, and R. C. Murray (eds.), Dolomitization and limestone diagenesis, Soc. Econ. Paleontologists and Mineralogists Spec. Pub., No. 13, p. 112-123. Shinn, E.A., Lloyd, R.M. and Ginsburg, R.N., 1969. Anatomy of a modern carbonate tidal- flat, Andros Island, Bahamas. Journal of Sedimentary Research, 39(3), pp.1202-1228. Sorensen, R.M., 2005. Basic coastal engineering (Vol. 10). Springer Science & Business Media. Stewart, S.R., 2017. National hurricane center tropical cyclone report: Hurricane Matthew. AL14016. National Oceanographic and Atmospheric Administration (NOAA). Stoddart, D.R., 1971. Terrestrial fauna of Diego Garcia and other Chagos atolls. Atoll Research Bulletin. No. 149: 163-170 Storlazzi, C.D., Elias, E.P.L., Berkowitz, P., 2015. Many atolls may be uninhabitable within decades due to climate change. Sci. Rep. 5, 14546. Taylor, K.H., Purkis, S.J., 2012. Evidence for the southward migration of mud banks in Florida Bay. Mar. Geol. 311–314, pp. 52–56. Thieler, E.R., Himmelstoss, E.A., Zichichi, J.L. and Ergul, A., 2009. The Digital Shoreline Analysis System (DSAS) version 4.0-an ArcGIS extension for calculating shoreline change (No. 2008-1278). US Geological Survey. Tuck, M.E., Kench, P.S., Ford, M.R. and Masselink, G., 2019. Physical modelling of the response of reef islands to sea-level rise. Geology, 47(9), pp. 803-806. Van Woesik, R. and Cacciapaglia, C.W., 2018. Keeping up with Sea-Level Rise: Carbonate Production Rates in Palau and Yap, western Pacific Ocean. PloS one, 13(5), p.e0197077. Webb, A.P., Kench, P.S., 2010. The dynamic response of reef islands to sea-level rise: evidence from multidecadal analysis of island change in the Central Pacific. Glob. Plan- et. Chang. 72, pp. 234–246. Woodroffe, C.D., 2005. Late Quaternary sea-level highstands in the central and Eastern Indian Ocean: a review. Glob. Planet. Chang. 49, pp. 121–138. Woodroffe, C.D., 2008. Reef-island topography and the vulnerability of atolls to sea-level rise. Global and Planetary Change, 62(1-2), pp.77-96. Yates, M.L., Le Cozannet, G., Garcin, M., Salaï, E. and Walker, P., 2013. Multidecadal atoll shoreline change on Manihi and Manuae, French Polynesia. Journal of Coastal Research, 29(4), pp.870