mapping bathymetry and depositional facies on...

Mapping bathymetry and depositional facies on Great BahamaBank

PAUL M. (MITCH) HARRIS* ,1 , SAM J. PURKIS† , JAMES ELLIS‡ , PETER K. SWART§and JOHN J. G. REIJMER¶*Chevron Energy Technology Company, 1500 Louisiana Street, #32048, Houston, Texas, 77002, USA(E-mail: [email protected])†National Coral Reef Institute, Nova Southeastern University, Dania Beach, Florida, 33004, USA‡Ellis GeoSpatial, Walnut Creek, California, 94598, USA§University of Miami, CSL – Center for Carbonate Research and Stable Isotope Laboratory, Departmentof Geosciences, Miami, Florida, 33149, USA¶Vrije Universiteit Amsterdam, Faculty of Earth and Life Sciences, Sedimentology and Marine GeologyGroup, Amsterdam, The Netherlands

Associate Editor – Giovanna Della Porta

ABSTRACT

Satellite imagery and an extensive set of water-depth measurements have beenused to map and critically evaluate the magnitude and patterns of bathymetryacross Great Bahama Bank. Descriptions of previously collected sediment sam-ples were combined with satellite imagery to map and refine the interpreteddistribution of surficial carbonate sediments (depositional facies). Data revealthat 60% of Great Bahama Bank lies in 5 m or less of water. The deep portionoccurs mainly in a generally east–west trending area in the southern portion ofthe platform. The re-evaluation of the facies reveals that Great Bahama Bank isessentially a very grainy platform with muddier accumulations primarily inthe lee of Andros Island. This area of Great Bahama Bank also experiencescurrents related to an excursion of the Florida Current onto the platform top;possibly enhancing sediment mud production through the generation of whit-ings. Sediment equivalents to mudstones, wackestones and mud-rich pack-stones cover 8%, 5% and 14%, respectively, of the platform top, whereassediment equivalents to mud-poor packstones, grainstones and rudstonesaccount for 20%, 45% and 3% of the surface area. Boundstones (reefs) werenot specifically mapped in this study due to the resolution of the mapping.There is a poor relationship between the occurrence of the depositional textureand water depth, in that the grainier sediment types are abundant across thefull range of water depths. The most abrupt lateral facies changes portrayed onthe facies maps are observed leeward of islands, areas which also hold thehighest diversity in facies type. The majority of the islands on the platformalign with the north-west/south-east strike of the platform margin and theseislands, in turn, exert control on the shape and orientation of facies belts thatdevelop in proximity to them. For this reason, regions of the platform that con-tain principal islands host facies belts that align with the principal axis of theplatform, whereas for regions lacking islands, the facies belts adopt an east–west trend consistent with prevailing winds and currents. There is a cleartrend that the wide southern portion of the platform hosts the most continuousexpanses of grain-rich sediments.

Keywords Currents, GIS, Great Bahama Bank, facies, Landsat, water depth,whitings.

1Present address: University of Miami, CSL - Center for Carbonate Research, Miami, Florida, 33149, USA

1© 2014 The Authors. Sedimentology © 2014 International Association of Sedimentologists

Sedimentology (2014) doi: 10.1111/sed.12159

INTRODUCTION

Great Bahama Bank (GBB) is ‘the’ modern exam-ple of a flat-topped, isolated carbonate platform.It is the major modern location of non-skeletalcarbonate deposition and, as such, GBB standsbehind much of the understanding of modernprocesses of carbonate sedimentation and is thebasis for geological models that commonly areused to illustrate depositional facies variationsand serve as reservoir analogues. New andrefined interpretive maps showing the variationin bathymetry and distribution of surficialcarbonate sediments (depositional facies) acrossGBB can only improve the utilization of this car-bonate platform as a natural laboratory for fur-ther investigation. Harris et al. (2014) describeand illustrate the steps of building the bathy-metry and facies maps in detail. Herein: (i) thesteps for building the maps are summarized; (ii)the applications of the new maps are empha-sized by focusing on several quantitative interro-gation approaches analyzing the water depthand facies relationships; and (iii) the aspects ofthe hydrodynamic controls over the facies pat-terns are discussed.

Previous mapping

The earliest understanding of the spatial distri-bution of surficial sediments across the GBBwas provided by Illing (1954), Newell et al.(1959), Purdy (1963a,b) and Traverse & Gins-burg (1966). These maps, along with those ofBall (1967) and Enos (1974), remain widelyused and were generally based on <50 samplingstations along a series of transects crossing thebank. Samples were collected without modernprecision positioning systems. In addition, grabsamplers were used, which may have led to aloss of the finer fraction of the sediments. Thelarge area of GBB (>100 000 km2) probablyintroduced additional sampling challenges.The distribution of surface sediments and

their geochemistry was re-evaluated on the wes-tern portion of GBB by Reijmer et al. (2009)and Swart et al. (2009) using 275 GPS-locatedsamples with a goal of quantifying the typesof sediments present in conjunction with minera-logy, grain-size distribution, sediment composi-tion and stable isotope geochemistry. Kaczmareket al. (2010) generated surficial sediment mapsfor several isolated carbonate platforms, includ-ing the entirety of GBB, by applying statistics-based unsupervised classifications to Landsat

Enhanced Thematic Mapper Plus (ETM+) multi-spectral satellite scenes. Landsat ETM+ pixelsare 30 m 9 30 m, which is an appropriate reso-lution to capture the broad scale of sedimento-logical variability atop the platform. Kaczmareket al. (2010) subsequently validated theirresults with sediment data to create geologicalfacies maps; in the case of GBB, these authorsused 14 samples provided as mapping con-straints.

Objectives of study

Although the number of samples used by Reij-mer et al. (2009) to characterize the surfacesediments represented a significant improve-ment over sampling by the early workers, theirmap only considers the western portion of theGBB, the same general areas as dealt with byPurdy (1963a,b) and Traverse & Ginsburg(1966) (Fig. 1). In order to attempt to improveresolution of bathymetry and facies across theentirety of GBB, first an extensive set of waterdepth measurements (n = 5731) was used inconjunction with Landsat Thematic Mapper(TM) and ETM+ imagery to create a digital ele-vation model (DEM) showing water-depth vari-ation. Facies data from the Reijmer et al. (2009)samples (n = 275) and additional groundtruthedsamples (n = 21) and Landsat imagery werethen compiled into ArcGIS and analyzed usingthe software eCognition (version 8.9, TrimbleInc., Munich, Germany) to develop a robustdepositional facies map. The facies map pro-vides a better understanding of facies distribu-tion across a flat-topped carbonate platformthan previous maps, and a comparison betweenthe bathymetry and facies from the maps leadsto a reconsideration of the water depth – faciesdoctrine, as recently considered by Purkis et al.(2014a) in the Red Sea.

METHODS

Landsat mosiac

Thirty images in thirteen footprints wereacquired by the Landsat 5 TM and 7 ETM+ sen-sors between 1985 and 2011 covering GBB.Because cloud cover, cloud shadows and sun-glint can degrade the quality of individualscenes, for many areas two to four Landsatimages acquired on different dates were evalu-ated. In several parts of the platform, more than

© 2014 The Authors. Sedimentology © 2014 International Association of Sedimentologists, Sedimentology

2 P. M. Harris et al.

A B

D E

C

Fig. 1. A sampling of previous facies maps of GBB. Traverse & Ginsburg (1966) (A), Purdy (1963a,b) (B) and Enos(1974) (D) are maps created without remote sensing imagery and precise positioning of samples. Reijmer et al.(2009) (C) and Kaczmarek et al. (2010) (E), like this study, rely in part on the interpretation of satellite imageryand are calibrated to differing degrees by GPS-located seabed samples.


Mapping bathymetry and facies on Great Bahama Bank 3

one image had to be used to provide a cloud-freescene of the islands and platform edge to ensuregood control for the bathymetric model. The thir-teen scenes were stitched together to create aplatform-wide mosaic. This 30 m mosaic is17 300 9 17 180 pixels in size – or contains294 410 000 pixels to manage, with each pixelcontaining three layers of grey-scale imagery(reflected blue, green and red light). This mosaicwas re-sampled to 100 m pixels so that the dataset covering GBB was more computationallymanageable for mapping of the entire platform –the resulting mosaic is 5244 9 5208 pixels insize and contains 27 268 000 pixels. Pixel resolu-tion of 100 m means that each 100 km2 of plat-form-top imagery is captured by 10 000 pixels.

Digital elevation model

A digital elevation model (DEM) showing bathy-metric variation across GBB was created by pro-cessing the satellite imagery with water-depthmeasurements from published navigation chartsand from field measurements with GIS software(Fig. 2). All water-depth measurements were cor-rected to Lowest Astronomical Tide and thereforethe DEM is also tendered in this vertical datum.The tidal range on GBB varies from ca 1 m at themargin of the platform (Wilson & Roberts, 1992,1995) decreasing to 0�3 m at the Three Creeks por-tion of the Andros tidal flats (Hardie & Garrett,1977). Over 5500 water-depth points were digi-tized from a published navigation chart – theExplorer Chart 1 named ‘The Bahama Islands’(© 2010 Lewis Offshore Ltd) (Fig. 2A). This papermap, which is published at a scale of 1 : 1 000 000,was scanned and georeferenced into a GIS usingthe latitude/longitude grid on the map. An addi-tional 210 soundings (Reijmer et al., 2009) and 21measurements collected by the National CoralReef Institute were also included. The edge of theGBB platform was interpreted visually and digi-tized (with reference to the soundings and pub-lished navigation charts) as a 30 m contour fromband 1 (reflected blue light) of the Landsat mosaic(Fig. 2B). Shorelines of the 1427 islands that sitatop the GBB were derived from the shortwave-infrared (SWIR) band 5 of the Landsat mosaic(Fig. 2C).The 30 m bathymetric contour and the island

shorelines were used to form the lower andupper boundaries, respectively, for the platformDEM in triangulated irregular network (TIN) for-mat (Fig. 2D). The TIN model was converted toa regular 150 m grid using the interpolated

distance weighting (IDW) method to facilitatevisualization (Fig. 2E). Resampling to 150 mprovides a grid of 3186 9 3209 or 10 224 000cells across the entire platform, and the bathy-metric control (soundings, contours and shore-lines) makes this a reasonable grid interval. Theindividual soundings at approximately six con-trol points per 100 km2 are not adequate for thebathymetry model; the contours and shorelineprovide much needed control. The satellitemosaic was resampled from 30 m to 100 m tosupport the facies mapping.Lastly, bathymetric surfaces created by Harris

et al. (2010, 2011) for the Exumas, Schoonersand Tongue of the Ocean (TOTO) sand bodieswere melded into the DEM (Fig. 2F) to yield thefinal bathymetric model (Fig. 3). The 30 mSchooner, Exumas, and TOTO sand body DEMswere re-sampled to the 150 m grid because thegoal was a manageable, platform-wide bathymet-ric DEM that would support visualization of the100 m resolution facies map.

Depositional facies map

The extensive set of bottom sediment samplescollected by Reijmer et al. (2009) was critical tothe current mapping study, because the faciesmap for the north-western portion of GBBmatches the sediment description at each sam-ple location. These samples were collected usinga Shipek sampler; the design of this samplereffectively prevents muds from being lost duringretrieval and thus differs from the grab samplerused in previous studies as discussed by Reij-mer et al., 2009. The dried samples were wetsieved to separate the coarse (>63 lm) from thefine fraction (<63 lm); see Reijmer et al. (2009)for a detailed description of the methods used toclassify the sediments. The sediment types(depositional textures) were modified from thatof Reijmer et al. (2009) for the current mappingstudy to be more consistent with the classifica-tion scheme of carbonate rock types proposed byDunham (1962) and Lucia (1995) (Table 1).Although there is a suggestion that a particularsediment type from the GBB samples will ulti-mately become the ‘equivalent’ rock type in theclassification, this interpretation is fairlystraightforward and justified. The distinctionbetween grains and mud matrix observed in thebottom sediment samples and assessment oftheir relative importance (grain-supported ormud-supported) is not fundamentally differentto what is done when describing a rock sample.



An object-oriented approach was adopted fordelineating surficial sediment bodies in theLandsat mosaic. This approach contrasts withKaczmarek et al. (2010) who followed Rankey(2002) and employed a pixel-based unsupervised

classifier to partition the platform-top of GBB intospectral classes which, through comparison witha small number of sediment samples, wereclumped into a final suite of classes deemed torepresent the distribution of seabed sediments.

A B

C D

Fig. 2. Summary of workflow for building the water-depth map of GBB: (A) Map showing locations of digitizedwater-depth soundings (932 shown here, 5510 in total) from Explorer Chart 1 ‘The Bahama Islands’ with others@2010 Lewis Offshore Ltd. (B) The edge of the platform was visually interpreted and digitized (with reference tosoundings) as a 30 m contour from the blue light Landsat TM band 1. (C) Shorelines of islands were derived fromthe SWIR Landsat band 5; water is black and land is shades of grey. (D) The bathymetry depth model used the tri-angular irregular network (TIN) interpolation method to integrate the 5723 soundings, edge of platform contour,interpreted intermediate contours and island shorelines.



The principal disadvantage of the unsupervisedclassification approach in a submerged settingsuch as the GBB, is that since light in the visiblespectrum is so rapidly attenuated by water, bathy-metric variations account for the majority ofspectral variation within the remote sensingimagery, rendering the platform-top sedimento-logical differences more difficult to differentiatedue to spectral differences (Purkis & Pasterkamp,2004; Purkis, 2005).In contrast to pixel-based classification meth-

ods, object-oriented image analysis, the strategyused to produce the facies maps in the presentstudy, segments satellite data into landscapeobjects that have geologically meaningfulshapes, and classifies the objects across spatial,spectral and textural scales (Blaschke & Hay,2001; Wang et al., 2004). An object-orientedclassification was employed to delineate deposi-tional facies ‘bodies’, interpreted to be distinctpatches of uniform surface sediment cover.Because of the flexibility afforded by includingnon-spectral attributes of the imagery (for exam-ple, textural, spatial and contextual information)into the classification workflow, object-orientedmethods have been shown to yield significantlyimproved accuracy over traditional pixel-basedimage analysis techniques (Kelly & Tuxen, 2009;Purkis & Klemas, 2011; Purkis et al., 2014a,b).

The software used for mapping in this study,eCognition (v. 8.9, Trimble Inc.), tenders a suiteof object-oriented image analysis algorithmshaving particular utility for creating thematicmaps from remote sensing data, including faciesmaps of modern carbonate platforms (Purkiset al., 2012, 2014a,b; Schlager & Purkis, 2013).For the portion of the GBB lying west of AndrosIsland, the facies map was generated by pairingthe extensive set of GPS-constrained field obser-vations and samples (n = 275) from Reijmeret al. (2009) and additional samples (n = 21) col-lected immediately north and south of Androsby the National Coral Reef Institute with object-based interpretations of the Landsat imagerywithin eCognition. The field observations strad-dle all of the benthic habitats found atop theGBB and therefore sediment-character could beadequately appraised, even for areas colonizedby submerged aquatic vegetation such as sea-grass. The Reijmer et al. (2009) facies sampleswere collected over an area of ca 35 000 km2, fora control density of approximately one controlpoint per 100 km2. Statistics pertaining to thespectral and textural properties of the Landsatimagery corresponding to the facies types areextracted at points where the samples provide anunequivocal determination of facies type, andthese statistics are used to drive a preliminary

E F

Fig. 2. (E) The TIN model was converted to a regular 150 m grid using the interpolated distance weighting (IDW)method to facilitate visualization. (F) Higher resolution (30 m) bathymetric DEMs of the Exumas, Schooners andTOTO sand bodies from Harris et al. (2010, 2011) were resampled to a 150 m grid for integration with platform DEM.



segmentation of the imagery into landscapeobjects using eCognition (Fig. 4A to C). With ref-erence to the seabed data, all landscape objectsare assigned to a facies category on the basis oftheir spectral and textural signatures (Fig. 4D).Next, a filter is applied to remove redundantdivisions between objects – i.e. those divisionsseparating two objects of the same facies category(Fig. 4E). A smoothing filter is used to refine theremaining landscape objects into a final faciesmap composed of polygons (Fig. 4F). The deposi-tional facies map honours all of the groundtruthdata on this portion of the platform; it is therefore100% accurate relative to the 296 samples.For the remainder of the platform, where such

rigorous ground-control is lacking, the Landsat

Fig. 3. The water-depth map (bathymetric DEM) for the GBB based on Landsat mosaic and soundings. Stepstaken to develop the map are described in the text and illustrated in Fig. 2. Key localities cited in the text areidentified on the map.

Table 1. Comparison of depositional texture (sedi-ment type) schemes used by Reijmer et al. (2009) andmodified for this study.

Reijmer et al. (2009) facies This study

1 = mudstone1�5 = mud-rich wackestone

1 = mudstone

2 = wackestone 2 = wackestone

2�5 = mud-rich packstone3 = packstone

3 = mud-rich packstone

3�5 = mud-rich grainstone 4 = mud-poor packstone

4 = grainstone 5 = grainstone

5 = rudstone 6 = rudstone



imagery was also segmented into lithotopes –interpreted to be distinct bodies of uniform sedi-ment – using a combination of edge-detection,spectral and textural analysis and manual edit-ing. The latter step was accomplished within

eCognition and used in the rare cases where thesoftware clearly miss-picked the boundary of asediment body, as occasionally caused by spec-tral artefacts in the Landsat imagery, such as sunglint off of the sea surface or an inconsistency in

A

A

B

C

D

E

F

Fig. 4. Workflow for the creation of the facies map: (A) GPS-located seabed samples are assembled within a GISatop the Landsat imagery; (B) shorelines of islands are identified in the SWIR Landsat band 5 and emergent areasare masked; (C) Statistics pertaining to the spectral and textural properties of the Landsat imagery correspondingto the facies types are extracted at points where the samples provide an unequivocal determination of facies type,and are used to drive a preliminary segmentation of the imagery into landscape objects using eCognition; (D)With reference to the seabed data, all landscape objects are assigned to a facies category on the basis of theirspectral and textural signatures; (E) A filter is applied to remove redundant divisions between objects; (F) asmoothing filter is used to refine the remaining landscape objects and the final facies map composed of polygons.



the seam separating two images in the mosaic.For this ‘unconditioned’ eastern portion of GBB,the spectral and textural statistics from the 296points for which sediment type was unequivo-cally known on the western limb of the platformwere used to guide the eCognition segmentationof the Landsat imagery into objects. Under thereasonable assumption that the sedimentologicaldiversity is equal for the eastern and westernlimbs of the GBB, the knowledge from the sedi-ment samples was therefore expanded to aidclassification in unsampled areas of the platformusing the quantitative and objective criteriaencoded within the spectral and textural analysis

of the Landsat mosaic. Figure 5 was then deve-loped by assigning lithotopes to facies classes onthe basis of lessons derived from the portion ofthe platform for which there was rigorous condi-tioning from the bottom sediment samples.

Analyzing water depth and facies

To identify trends in the arrangement of sedimenttypes with respect to water depth, a standard isneeded against which to judge the tendency for aparticular facies to associate with a specific waterdepth. To this end, the workflow adopted by Ran-key (2004) on the South Florida shelf and Purkis

Fig. 5. Map of depositional facies (sediment types) for GBB based on sea floor sample data compiled into ArcGIS,integrated with the Landsat mosaic and analyzed with eCognition. Steps taken to develop the map are describedin the text and are shown on Fig. 4. The facies map is superimposed upon a simplified version of the bathymetricDEM for GBB.



et al. (2014a) in the Red Sea is followed, and themaximum entropy concept is employed wherebydivergence from a state of disorder is statisticallyassessed. The end-members to this approach areperfect determinism – one water depth, one sedi-ment type – and randomness, where the deptharrangement of a number of sediment typesexhibits a state of maximum disorder (waterdepth and facies are independent).The Shannon evenness index was used to

examine facies diversity (substitutability) acrossthe range of water depths recorded atop theplatform (Shannon, 1948; Rankey, 2004). Thecalculation proceeds as follows. Given a water-depth range in which there exist n possiblefacies classes, with proportions pi,. . .,pn withinthat water-depth range, the Shannon evenness(E), also termed the ‘deterministic component’for each water depth is calculated as:

E ¼ 1��Pn

i¼1

pi � lnpi

lnðnÞ ð1Þ

In order to implement Equation 1, the bathyme-try was partitioned into 1 m increments span-ning 0 to 16 m. Erroneous interpretation of thepartitioning of facies by depth across incre-ments with few observations was precluded bynot computing E for depth intervals containingfewer than 1000 facies observations. E scalesfrom 0 to 100%, with values near unity indicat-ing that an increment of water depth is notdiverse, but instead dominated by one faciesclass. In this situation, there is a more deter-ministic relation; given a water depth, the sedi-ment type present can be predicted withconfidence (Rankey, 2004). Under such condi-tions, by looking at a map of facies, a map of depthcould be inferred, or vice versa. E equals zerofor the situation where a depth or wave height binis occupied by an equal proportion of all faciesclasses – the case of maximum entropy and onefor which knowledge of water depth carries theleast predictive power for facies class.Between zero and 100%, E is proportional to

the percentage that uncertainty has beenreduced from the maximum. For example, asdescribed by Rankey (2004, p. 3): “for a givenwater depth, a value of E = 20% means that theobserved uncertainty in class occurrence hasbeen reduced 20% relative to the maximumpossible entropy and, conversely, that there is a20% deterministic or predictable component, asconstrained by water depth”.

RESULTS

Platform-scale characterization

Great Bahama Bank covers an area of over103 000 km2 (590 km north–south and 160 kmeast–west) and the margin of this immenseplatform extends for 3088 km in length. Addi-tional notable platform-scale attributes are: (i) ifthe deep re-entrants of TOTO and Exuma Soundwere filled, the platform would be oval in shapewith a principal axes ratio of 1�7; (ii) the areacovered by the two re-entrants is broadly equiva-lent at 15 200 km2 and 12 000 km2, respectively;and (iii) the combined areas of the re-entrants,27 200 km2, is 25% of the total platform area.Figure 6 further characterizes the margin of GBBwith respect to its orientation, showing that ca60% of the platform margin is north–south ori-entated (1835 km), 20% is east–west orientated(640 km) and the remaining 20% has obliqueorientations.

Digital elevation model

The maximum variation in depth-elevation overthe vast GBB platform extends from the 30 mbathymetry contour to the highest Pleistoceneaeolianite ridge of 63 m on Cat Island; but,because islands occupy only 8%, or 8000 km2,of GBB, the bathymetric variation is emphasizedhere. Sixty percent, or 61 400 km2, of the sub-merged portion of GBB lies in 5 m or less ofwater (shown in tan on Fig. 7). This includesareas where accommodation space has beenlocally filled, for example, the sand bodies inthe southern TOTO (Ball, 1967; Palmer, 1979;Harris et al., 2010, 2011), Schooners Cays (Ball,1967; Dravis, 1977; Rankey et al., 2008; Harriset al., 2010, 2011; Sparks & Rankey, 2013), Exu-mas (Harris, 2010; Harris et al., 2010, 2011),Joulters Cays (Harris, 1979) and Cat Cay (Ball,1967; Cruz, 2008) areas. Also included are vastareas of the northern portion of GBB surround-ing Andros Island and the New Providence Plat-form to the west of the Exuma Islands. The 40%of GBB lying in greater than 5 m of water(shown in blue on Fig. 7) occurs mainly in agenerally east–west trending area toward thesouthern end of the platform.

Depositional facies map

The depositional facies map reveals the GBB tobe essentially a very grainy platform with



Fig. 6. Simplified, shaded-reliefversion of the bathymetric DEM ofGBB to emphasize platform-scalecharacterization possibilities; theplatform margins are characterizedwith respect to their orientation.

Fig. 7. A version of the bathymetricmap (Fig. 3) coloured to highlightshallower (<5 m water depth shownin tan) and deeper (>5 m waterdepth shown in blue) portions ofGBB.



muddier accumulations only in the lee ofisland barriers (Fig. 8), the most notable ofwhich is Andros Island, the largest island onGBB covering an area of 4000 km2. Sedimentequivalents to mudstones, wackestones andmud-rich packstones cover 8%, 5% and 14%,respectively, of the GBB platform top (Fig. 5).In contrast, sediment equivalents to mud-poorpackstones, grainstones and rudstones accountfor 20%, 45% and 3%, respectively (Fig. 5).Boundstones (reefs) were not specificallymapped due to the resolution at which thefacies mapping was done. Of the 45% of theplatform-top classified as grainstone, only 3% iscomposed of ‘high-energy’ deposits characte-rized by the development of sandbar complexeswherein variably thick, cross-bedded ooid sandsare found. The remaining mapped grainstonesare subtidal sand sheets, containing more uni-formly thin sands with variable grain types[superficial ooids, peloids, skeletal grains andaggregate (grapestone) grains] and variable sedi-mentary structures (bioturbation, burrowing andlocal laminations).The diversity and size of facies bodies shown

are broadly the same on the eastern and westernlimb of the GBB platform, although the narrower

eastern New Providence Platform hosts a higherprevalence of high-energy grainstones (Fig. 5).The most abrupt lateral facies changes as por-trayed on the maps are observed leeward ofislands, areas which also hold the highest diver-sity in facies type. Higher resolution mappingwould show rapid facies changes seaward ofislands in the narrow but variable outermostportion of the platform, but this variability isnot captured at the mapping scale used. Thefacies map also reveals the northern half of theplatform to host a more heterogeneous faciesmosaic than the southern half, a difference prob-ably related to the greater prevalence of islands.There is a clear trend that the widest east–westportion of the platform, which lies to the southof TOTO and lacks islands, hosts the most con-tinuous expanses of grain-supported sediments(Fig. 9). Further, the breadth of grainy fairwayson the platform top is double that of the mud-rich areas, grainy lithologies are larger on thesouthern platform (in deeper water), and grainysedimentary bodies are highly interconnectedwhereas mud-rich bodies are generally smallerand more isolated (Fig. 9).The orientation of depositional facies from

Fig. 5 can be investigated with regard to island

Fig. 8. A version of the facies map(Fig. 5) coloured to emphasizemuddier (mudstone, wackestone,and mud-rich packstone shown intan) and grainier (mud-poorpackstone, grainstone, and rudstoneshown in blue) portions of GBB.



occurrence and orientation of the platform mar-gin. In order to compare patterns of facies align-ment across the GBB, the platform waspartitioned into four regions on the basis of theoccurrence of islands (Fig. 10). For regions 1 and3 of GBB with few and small islands only, the ori-entation of facies belts is unrelated to the orienta-tion of islands. In contrast, for regions 2 and 4 ofGBB with numerous islands, there is consistencyin orientation between islands with areas exceed-ing 3000 km2 and the largest facies belts. If theorientations of the mud-rich facies are consideredon their own merit, Fig. 10 also shows the influ-ence of islands. For example, the orientation ofthe muddy facies belts is predominantly east–west in regions 1 and 3, whereas the sedimentbodies are island-aligned in regions 2 and 4.

DISCUSSION

Comparison to previous facies maps of theGreat Bahama Bank

The first map of GBB to use the Dunham classifi-cation scheme (Dunham, 1962) portrayed the

relative importance of the distribution of grainsand mud (= matrix) – or depositional texture ofDunham (1962) – across the platform (Enos,1974) (Fig. 1D). The Reijmer et al. (2009)(Fig. 1C) and Kaczmarek et al. (2010) (Fig. 1E)maps also use the Dunham terms as a basis formapping, as in the present study. Of particularimportance is the detailed and generalized dis-tribution of muddier (mudstones, wackestonesand mud-rich packstones) and grainier litholo-gies (mud-poor packstones, grainstones and rud-stones) shown on Figs 8 and 5 relative to that ofprevious maps. A comparison of the early maps(Fig. 1) with Fig. 5 allows assessment of simila-rities, differences and improvements. Two com-mon parameters that can be used to compare thevarious facies maps of GBB are the proportion ofeach facies and the number of distinct mapped‘bodies’ (Table 2). Comparison of the formerhighlights the differences between the sedimen-tological interpretations of the platform, whereasboth parameters are convenient metrics of mapcomplexity.The early maps by Newell et al. (1959),

Purdy (1963a,b), Traverse & Ginsburg (1966)and Ball (1967) covered only portions of GBB

Fig. 9. The spatial facies patterns from Fig. 5 are summarized into planform width maps for grainier sediment(left) and muddier sediment (right) facies bodies. Hotter colours demark wider, more laterally expansive, faciesbodies. Grainy bodies are shown to be both more laterally expansive and interconnected than their muddy coun-terparts.



Fig.10.Analysisoffaciesorientationswherein

GBB

issu

bdividedinto

regionsandthealignmentofdepositionalfaciesfrom

Fig.5are

investigated

with

regard

toislandoccurrenceandplatform

margin

orientation.Faciesbeltsorientparallelto

theplatform

margin

whenislandsare

present.



and were generated without the benefit ofremote sensing imagery. The map of Enos(1974), arguably the most widely used map ofGBB, covers the entirety of GBB but is also notbased on remote sensing imagery. Of the earliermaps, only the Purdy (1963a,b) map shows thelocations of bottom samples which were usedas control points, but the accurate positioningof sample locations is a concern. All of thesemaps show that: (i) skeletal sediments (grain-stones and rudstones) and local ooid sands(grainstones) occur mainly along the platformmargin; (ii) non-skeletal grains (mainly pelletsor peloids) dominate the interior of the plat-form; and (iii) very fine-grained sediments pre-dominate in the platform interior whereprotected by a topographic barrier, for example,‘pelletoidal sands with lime mud matrix andlime muds’ (Traverse & Ginsburg, 1966),‘muddy sand and mud’ (Newell et al., 1959),‘pellet-mud facies’ and ‘mud facies’ (Purdy,1963a) and ‘pelletoidal packstones’ and ‘pellet-oidal wackestones’ (Enos, 1974).Reassuringly, despite different modes of con-

struction and ground-control, all of the mapsshow comparable areal coverage of the commonfacies classes: grainstones dominate, packstoneis the next most prevalent texture, followed bywackestone, and mudstone and rudstone are

each judged to inhabit <10% of the platform-top(Table 2). Some differences exist, however, inthe details of the facies types as shown on thevarious maps. For example, Traverse & Ginsburg(1966) and Enos (1974) show that the sedimentsin the platform interior mainly contain pellets,peloids or pelletoids. Newell et al. (1959) andPurdy (1963a,b), however, describe sedimentscontaining a larger percentage of grapestones.Enos (1974) also described the occurrence ofgrapestones without showing them on the mapas a particular facies type.All of the GBB facies maps show the variabi-

lity in mud-dominated facies occurring in thewestern portion in the lee of Andros Island, butthe details of this variability are shown only onthe maps of Reijmer et al. (2009), Kaczmareket al. (2010) and this study (Fig. 5). The tech-niques used by Kaczmarek et al. (2010) and inthis study to construct a GBB facies map differin approach and detail. An important differencebetween the two studies is that Fig. 5 preciselyhonours all of the sample points of Reijmer et al.(2009) as a key step in the mapping workflow,while also respecting the distribution of sedi-ment ‘bodies’ that are visible clearly in the satel-lite imagery and captured by the numericinterpretation of the spectral and textural contentof the Landsat data using eCognition. Although

Table 2. Comparison of amounts of sediment types mapped by previous workers and in this study. Because ofdifferences in the scheme for each map considered, some facies categories had to be lumped to provide a mean-ingful comparison. For instance, mudstones mapped by Traverse & Ginsburg (1966) were lumped into the wacke-stone category (Wacke) as were mud-rich packstone, whereas mud-poor packstone was lumped to grainstone(Grnstn). NA = data unavailable.

Traverse & Ginsburg(1966) Enos (1974)

Reijmeret al. (2009)

Kaczmareket al. (2010)

Thisstudy

Mudstone % GBB† Lumped to Wacke Not mapped 9% 0% 8%# bodies‡ Lumped to Wacke Not mapped NA 0 2668

Wackestone % GBB 18% 8% 11% 7% 5%# bodies 1 6 NA 3 17

Mud-rich packstone % GBB Lumped to Wacke 14% 9% 24% 14%# bodies Lumped to Wacke 5 NA >500 56

Mud-poor packstone % GBB Lumped to Grnstn Not mapped 11% 23% 20%# bodies Lumped to Grnstn Not mapped NA >200 90

Grainstone % GBB 63% 78% 54% 46% 50%# bodies 1 2 NA >500 1160

Rudstone % GBB 18% Not mapped 0% 0% 3%# bodies 1 Not mapped NA 0 602

†Percentage of occupancy of a given facies atop the Great Bahama Bank. ‡Number of distinct facies bodies inmap.



all of the maps show generally concentric faciesbelts that extend roughly parallel to Andros,there are differences in the maps for the areas tothe west and north-west of Andros Island withFig. 5 generally showing more detail.A comparison of the Kaczmarek et al. (2010)

map and Fig. 5 for the remainder of GBB showsgenerally similar facies patterns. The exceptionis that Fig. 5 shows more detail with regard tothe patchiness of packstones within an other-wise grainstone-dominated GBB (Table 2). TheKaczmarek et al. (2010) map shows a greaterexpanse of packstone atop the south-westernportion of the GBB than is indicated in thegroundtruthed maps of Purdy (1963a,b), Reijmeret al. (2009) and the present study. Nevertheless,the proportion (Table 2) and the pattern (Fig. 5)of the grain-dominated nature of GBB agree withthose reported by Kaczmarek et al. (2010) andEnos (1974).

Facies water-depth relations on GreatBahama Bank

Bathymetric patterns are important as a poten-tial primary control over flooding history, filling(accommodation) history and the resultant depo-sitional facies. The bathymetric DEM (Fig. 3)highlights the irregular filling of accommodationspace and graphically emphasizes the challengesdiscussed by Eberli (2013) of correlating deposi-tional cycles of variable thickness across aplatform.Reijmer et al. (2009) suggested from their

mapping a link between water depth and typesof sediment on the platform. Coarse-grained,mud-free facies occur close to the edge of theplatform, on relatively high areas within theinterior of the platform (for example, south-eastof Bimini), and in the more open southern partof the platform. Newell et al. (1959) describedtwo broad shoals running across the westernpart of GBB: the Bimini axis extending fromBimini to the northern tip of Andros Island, andthe Billy Island axis running westward from themiddle promontory of Andros Island. Theserelatively higher areas are evident also in theresults of this study (Figs 3 and 7). The Reijmeret al. (2009) map of mean grain-size (fig. 3A)reflects these topographic features, whereas thefacies maps of Figs 5 and 8 suggest a rela-tionship between the Bimini axis and facies butdo not reflect the Billy Island axis. The moreopen character of the southern part of GBB isclearly reflected in the coarser, mud-poor sedi-

ments on the Reijmer et al. (2009) maps (fig. 3Ato D). Expanding the view completely across thesouthern portion of GBB (Fig. 5) shows the pre-dominance of packstones as well as grainstonesin this portion of the platform.The relationship between sediment type and

water depth on GBB, as proposed by Reijmeret al. (2009), can be further investigated quanti-tatively using a combination of the bathymetricand facies maps. There is a poor relationshipbetween the occurrence of the seven mappedfacies types and water depth (Fig. 11A) in thatthe grainier sediment types are abundant acrossthe full range of water depths. Muddier sedi-ments also span a considerable depth range,but do not extend as deep as the grainier facies.The poor relationship between facies and waterdepth is not surprising, in that a combinationof factors, including tidal velocities, waveenergy and local topography, create the localphysical environment on the sea bottom whichcontrols sediment characteristics.Relations between Shannon evenness (E) and

depth reveal a subtler trend (Fig. 11B). Whilethe deterministic component is low (<40%) inwater depths less than 8 m, it rises in the deeperwater depths, reaching a maximum of 70% inthe 15 to 16 m depth interval. Therefore, withknowledge of depth in the 0 to 8 m range, faciescan only be predicted 40% better than the situa-tion of maximum entropy. Meanwhile, for thedeepest portions of the GBB, sediment type canbe predicted 70% better than the equiprobablesituation, with the reason for the high level ofpredictability easily attributed to the high preva-lence of grainstones in this depth range.

Tides, currents and wind as controls of faciespatterns

While a comparative analysis of the surficialsediment (facies) and bathymetric maps doesnot yield a fully satisfactory explanation for theobserved patterning of sediment textures atopthe GBB, consideration of prevailing tides, oceancurrents and winds provides useful insight. Thetidal range on GBB varies from ca 1 m at themargin of the platform (Wilson & Roberts, 1992,1995) decreasing to 0�3 m at the Three Creeksportion of the Andros tidal flats (Hardie &Garrett, 1977). For northern Little Bahama Bank,Reeder & Rankey (2009), Rankey & Reeder(2011) and Rankey & Doolittle (2012) noted thetides to vary between ca 0�75 m during neaptides and 1 m during spring tides. Flow veloci-



ties associated with these tides range from 40 to80 cm s�1 within different ooid shoal complexesof Little Bahama Bank (Rankey et al., 2006;Reeder & Rankey, 2009; Rankey & Reeder, 2011),and similar velocities were measured on GBB ina tidal inlet of the Exumas (Gonzalez & Eberli,1997) and a tidal channel of the Schooners Caysarea (Rankey et al., 2008). Areas of focused tidalflow often are associated with the formation ofooid bar belts, as extensively considered bynumerous workers, such as Ball (1967), Hine(1977), Harris (1979), Hine et al. (1981), Reeder& Rankey (2008, 2009), Harris (2010), Rankey &Reeder (2012) and Harris et al. (2011).Reijmer et al. (2009) discuss the ocean cur-

rents and circulation that potentially impact sed-iment formation and redistribution across GBB,concluding that water movement on GBB ismostly influenced by wind and tides, but is alsoinfluenced by waves and storms. The great lat-eral extent, shallow water nature and presenceof islands along the platform margin of GBB gen-erally diminish the importance of oceanic cur-rents on the platform top. Averaged oceansurface current data (OSCAR) for calendar year2013 reveal dominant water flow across the GBBto be in a westerly direction which is in accor-dance with the direction of prevailing winds forthe region (Fig. 12; OSCAR data downloaded on2 January 2014 from www.oscar.noaa.gov). These

OSCAR data have a 50 km spatial resolution, aretendered in near-real time and derived fromsatellite altimeter and scatterometer data (Lager-loef et al., 1999; Lumpkin et al., 2009). The dataare groundtruthed by drifter buoys where avai-lable.There are two important exceptions to the

trend of generally westerly flow of water acrossthe GBB. The first is a southerly, but low velo-city current that sweeps the New ProvidencePlatform, related to inflow of ocean water intothe northern reaches of the TOTO (Fig. 12). Aswould be anticipated from the direction of flow,facies bodies in this region of the platform alignnorth–south. The second exception is a clock-wise circulation on the platform-top related tothe passage of the Florida Current through theStraits of Florida. As noted by Purkis et al.(2014b), the Florida Current largely remainsconfined to the deep topography of the FloridaStraits, probably obeying vortices constraints. Atthe point of maximum constriction between theFlorida Peninsula and the northern GBB, how-ever, a sustained incursion of the Florida Cur-rent occurs, driving a clockwise circulation cell,which deflects westwards in the lee of AndrosIsland and exits the platform as a westerly flowfurther to the south (Fig. 12). It is relevant tonote that this sector of GBB, in the lee of And-ros, witnesses the highest platform-top current

A B

Fig. 11. (A) Stacked bar-graph reporting the distribution of facies type versus water depth. (B) X–Y plot furtherexamining this relationship statistically using entropy. Facies distributions on the GBB are not shown to be statis-tically differentiated or limited across the platform-top.



Fig.12.Summary

ofwind

and

currentdata

forGBB:wind

rose

showsdata

extracted

from

QuikSCAT

satelliteobservationsfortheperiod

January

2013

through

toJanuary

2014;and

OSCAR

altim

eter–scatterometercurrentdata

forthesameperiod

show

broad

trendsofwatermovement,

both

direction

and

intensity,ontheplatform

-top.LBB

=LittleBahamaBank;GBB

=GreatBahamaBank;CSB

=CaySalBank.



velocities reported by the OSCAR data and cor-responds to the greatest degree of facies hetero-geneity interpreted from the Landsat (Fig. 5).This platform current pattern seems to haveinfluenced the production of carbonate mud andits subsequent transport to the westward marginof GBB (see discussion in the next paragraph).For the rest of the platform, the alignment offacies bodies corresponds to the east to westacross-platform current flow that appears to bewind-driven. Seasonal atmospheric circulationpatterns strongly influence temperature andprecipitation patterns on GBB. Circulationresponds to changes in the position and strengthof high-pressure and low-pressure systems, theIntertropical Convergence Zone (ITCZ), the Sub-tropical Divergence Zone (STDZ) and the associ-ated trade wind belt (Reijmer et al., 2009).

Whitings and mud distribution

Although GBB is a grainy platform, the produc-tion and distribution of lime mud are importantto understand within the regional context. Thesource of the muds in the Bahamas is controver-sial, with one possible origin being the sponta-neous precipitation within the water columnknown as whitings (Black, 1933; Cloud, 1962;Shinn et al., 1989). Whitings, areas of sedimentladen water, have been postulated to resulteither from bottom sediments stirred up bycurrents, fish or other physical mechanisms(Broecker & Takahashi, 1966; Morse et al., 1984,2003), through the direct precipitation of CaCO3

(Black, 1933; Cloud, 1962; Shinn et al., 1989), orthrough induced precipitation of CaCO3 by thephotosynthetic activity of cyanobacteria (Rob-bins, 1992; Yates & Robbins, 1998) and fertilizedby the input of Fe from atmospheric dust (Swartet al., 2014). Proponents of the direct precipita-tion origin point to the difference in the crystalstructures of the aragonite from the whitings andbottom sediments (Loreau, 1982; Reid & MacIn-tyre, 2000), the absence of mechanisms wherebythe sediments could be stirred up (Shinn et al.,1989), the absence of sufficient quantities of cal-cium carbonate producing organisms on the plat-form surface (Steinen et al., 1988; Shinn et al.,1989; Robbins et al., 1997) and differences in theconcentration of strontium between biogenic andnon-biogenic carbonates (Milliman et al., 1993).A reworked origin is supported by the fact that theradiometric dating of the whitings indicates anolder age than Modern (Morse et al., 1984; Broec-ker et al., 2000) and by the absence of changes in

the carbonate alkalinity between the whitings andthe surrounding waters (Morse et al., 1984, 2003)which would be expected if direct precipitationoccurred. Some of the researchers who had previ-ously advocated a stirred up origin based onthe absence in alkalinity changes within and out-side the whitings (Morse et al., 1984, 2003), haveacknowledged that there is some direct precipita-tion component in whitings studied on LittleBahama Bank (Bustos-Serrano et al., 2009). Thislater finding supports the Shinn et al. (1989)proposition, that whitings are composed of mix-tures of stirred-up bottom sediments and directaragonite precipitate.The relation between whitings and patterns of

muddier sediment types can be refined usingFig. 13. Robbins et al. (1997) previously com-pared the spatial and temporal distribution andlongevity of 888 whitings on western GBB byanalyzing 69 astronaut and satellite imagestaken from 1965 to 1993. The goal of this assess-ment was to glean from the size and frequencydistribution of whitings a measure of carbonatemud production. Robbins et al. (1997, fig. 2)concluded from comparing the distributionof whitings with the GBB facies map ofPurdy (1963b) that most whitings occurredwithin the muddier facies in the central part ofthe platform and the pellet-mud facies in theeast-central part of the platform. Figure 13re-emphasizes this comparison and shows theinterplay between whitings and sediment typesin greater detail. In particular, it emphasizes thatwhile the whitings occur predominantly in theareas of GBB where there are muddier sedi-ments, frequently whitings are found over por-tions of the bank in which the bottom sedimentsare grainier. This, at least, precludes the possi-bility that the whitings were derived from stir-red up bottom sediments in those particularareas and further strengthens the argument ofShinn et al. (1989).The whitings tabulated by Robbins et al.

(1997) are concentrated in a non-random,north-west/south-east oriented pattern that iscentred in the north-central part of westernGBB (Fig. 13). Robbins et al. (1997) estimatedthe amount of mud that could be producedfrom the area of whitings and then comparedthis to an estimate of the amount of mud accu-mulated on the platform top from ‘isopach’maps of Purdy (1963b), and the amount of mudtransported off of the platform from an isopachmap of the Holocene upper slope highstandwedge deposit documented by Wilber et al.



(1990). The coring of this wedge demonstratesthat the initiation of the export of carbonatemud coincided with the flooding of the platformca 7200 yr BP (Roth & Reijmer, 2004, 2005). Itwas concluded by Robbins et al. (1997) that,even using their lowest estimate of whitings,mud production creates a larger volume of limemud than is currently observed on the platformtop and could conceivably have sourced all orpart of the off-platform mud wedge (Robbinset al., 1997). Using the facies map of Fig. 5instead of the Purdy (1963b) map would reducethe estimate of mud exported from the platformby ca 10%, but the conclusions regarding mudbudget would not substantially change.The formation of whitings may be related to

the flow of water onto GBB (Fig. 12). Figure 12

shows that strong currents are present, bringingnew water from the Straits of Florida towardsAndros Island. In this region workers havefound a loss of alkalinity reflecting the precipi-tation of CaCO3 (Broecker & Takahashi, 1966;Morse et al., 1984). The implication is thatthese currents represent a pathway with whichthe water on GBB is continually being renewed.Similarly, as the waters flow south and exitGBB they can carry a significant proportion ofthe precipitated sediments off of the bank to bedeposited in deeper water, leading to the pro-gradation of the platform (Eberli & Ginsburg,1987). It may be of further significance that thecurrents shown on Fig. 12 appear in the lee ofAndros, a portion of the platform traditionallybelieved to be sheltered from currents caused

Fig. 13. The locations of whitingscompiled by Robbins et al. (1997)superimposed on the depositionalfacies map (Fig. 5).



by winds (Traverse & Ginsburg, 1966). Itappears, therefore, that the high productivity ofmud in this area, may not be a result of thelower currents in the area, as suggested in pre-vious studies (Purdy, 1963a,b; Traverse & Gins-burg, 1966; Enos, 1974), but are a result of thehigher productivity caused by continual inputof new water onto GBB and perhaps fertiliza-tion by atmospheric dust (Swart et al., 2014).

CONCLUSIONS

New interpretive maps showing water-depth var-iation, surficial carbonate sediment (facies) dis-tribution, and current flow have been created forGreat Bahama Bank (GBB). Because this ‘classic’isolated carbonate platform stands behind manyof the models used to illustrate depositionalfacies variation, the maps should provide a bettergeneralized understanding of facies distributionacross a flat-topped platform and should resultin improved models. The methods employed todevelop the maps, which integrate water-depthdata, previously published and new sedimentdata, as well as satellite-derived textural, geomet-ric and contextual information, are more sophis-ticated than those used in previous studies. Themaps, which represent new hypotheses for theGBB, are compared to a number of previouslypublished maps for the GBB; based on the newmethods and are in complete agreement with allavailable sediment data, the new maps containless uncertainty than the previously publishedmaps.The orientation of margins of GBB, along with

other factors such as expanse of the platformand occurrence of islands, play a role ininfluencing how tide-energy, current-energy,wind-energy and wave-energy impact sedimentdistribution across the relatively shallow top ofGBB and probably complicate the simplifiednotion of a windward and leeward margin.Although islands occupy only 8% of GBB plat-form-top, they influence the facies textures, het-erogeneity in those textures and alignment ofthe facies belts that accumulate in their lee.The largest mud accumulations occur in the

lee of Andros, the largest island on the GBB,with the remainder of the platform dominatedby grainy textures. Facies belts sited in sectorsof the platform with islands align with thoseislands which, in turn, because they are prefe-rentially situated on the platform margin, alignwith the strike of the platform (north-west/

south-east). In contrast, facies belts that accumu-late in exposed sectors of the platform align inthe east–west direction of the prevailing wind,suggesting windward–leeward influence, and aremore grainy than sectors with abundant islands.While mud-rich sediments might be precipitated(whitings), islands evidently impact the deposi-tion of the muddy facies belts. Tidal currents,along with wind-generated and wave-generatedcurrents, would all be diffracted around theislands (and other bathymetrical features),modifying current velocities and redistributingsediments. The impact of currents on sedimentmud production is shown by the concurrenceof a Florida Strait derived gyre and the produc-tion maxima of whitings in the lee of AndrosIsland.Also probably related to the distribution of

islands on the platform, the northern half of theGBB hosts a more heterogeneous facies mosaicthan the southern half for which water flowatop the platform related to an incursion ofthe Florida Current on the GBB is implicated. Thenarrower eastern New Providence Platform,meanwhile, hosts a higher prevalence of high-energy grainstones than the western limb of theGBB and current flow is shown to be influencedby prevailing wind. A total of 60% of the GBBplatform-top lies in 5 m or less of water. Ofthe 40% situated at greater than 5 m depth, themajority is located in a generally east–west trend-ing area of the southern portion of the platform.With the exception of water depths exceeding10 m, which are dominated by grainstones, themapped facies are not significantly differentiatedor limited across the platform-top, and thereforeit is not possible to predict the occurrence of agiven texture on the basis of bathymetry alone.These results suggest that relations between waterdepth and facies need to be carefully considered.The new facies maps can serve as a template

for better characterizing the GBB at all scales,highlight future research areas where ‘ground-truthing’ is needed to further investigate faciespatterns, and facilitate better use of this isolatedcarbonate platform as an analogue for bothexploration-scale and reservoir-scale subsurfacefacies analysis.

ACKNOWLEDGEMENTS

We thank Chevron Energy Technology Companyfor support of this research and for permissionto publish.



Lewis Offshore Ltd (www.explorercharts.com)kindly granted permission to extract bathymetriccontrol from their published and copyrightedExplorer Chartbook navigation charts of theBahama Islands (©2010 Lewis Offshore Ltd). Con-trol points and contour lines derived from theirnavigation charts were critical to successfullymodelling the bathymetry of GBB. Collection ofthe samples which provided the groundtruthingfor this study was provided by the Sponsors ofthe Comparative Sedimentology Laboratory,GEOMAR-Kiel, the German Science Foundation(DFG) and the Sedimentology and MarineGeology Group of the VU University Amsterdam.ESRI ArcGIS software was used extensively –

the TIN model, conversion of the TIN to a gridformat, bathymetric contouring and perspectivevisualization was done with the 3D Analystextension. ITT ENVI and Intergraph ERDASImagine software were used to process the satel-lite imagery. ET Spatial Techniques software ET-GeoWizards and Data East Soft, LLC softwareXTools Pro were used to process vector files.Landsat Imagery is courtesy of US Geologic Sur-vey. The backdrop map in some GIS figures isfrom ESRI’s online basemap service (Sources:ESRI, GEBCO, NOAA, National Geographic, De-Lorme, NAVTEQ, Geonames.org. and other con-tributors). We thank Matthew Johnston forassistance compiling wind and OSCAR currentdata and Lotte Purkis for assistance with eCogni-tion.Sam Purkis was supported by the National

Coral Reef Institute, Nova Southeastern Univer-sity Oceanographic Center. Ship time to collectthe original samples was supported by the Flo-rida Institute of Oceanography, GEOMAR-Kiel,and the Comparative Sedimentology and StableIsotope Laboratories at the University of Miami.Bob Ginsburg is thanked for continued support.We sincerely thank reviewer Stephen E. Kacz-

marek, an anonymous reviewer, and AssociateEditor Giovanna Della Porta who providednumerous comments that substantially improvedthe manuscript.

REFERENCES

Ball, M.M. (1967) Carbonate sand bodies of Florida and the

Bahamas. J. Sed. Petrol., 37, 556–591.Black, M. (1933) The precipitation of calcium carbonate on

the Bahama Bank. Geol. Mag., 70, 455–466.Blaschke, T. and Hay, G. (2001) Object-oriented image

analysis and scale-space: theory and methods for modeling

and evaluating multi-scale landscape structure. Int. Arch.

Photogramm. Remote Sens. Spat. Inf. Sci., 34, 22–29.Broecker, W.S. and Takahashi, T. (1966) Calcium carbonate

precipitation on Bahama Banks. J. Geo. Res., 71, 1575–1602.Broecker, W.S., Sanyal, A. and Takahashi, T. (2000) The

origin of Bahamian whitings revisited. Geophys. Res. Lett.,27, 3759–3760.

Bustos-Serrano, H., Morse, J.W. and Millero, F.J. (2009) Theformation of whitings on the Little Bahama Bank. Mar.

Chem., 113, 1–8.Cloud, P.E. (1962) Environment of calcium carbonate

deposition west of Andros Island, Bahamas. US Geol.

Surv. Prof. Pap., 350, 1–138.Cruz, F.E. (2008) Processes, patterns, and petrophysical

heterogeneity of grainstone shoals at Ocean Cay, Western

Great Bahama Bank. PhD Dissertation, University of

Miami, Miami, FL.

Dravis, J.J. (1977) Rapid and widespread generation of recent

oolitic hardgrounds on a high energy Bahamain platform,

Eleuthera Bank, Bahamas. J. Sed. Petrol., 49, 195–208.Dunham, R.J. (1962) Classification of carbonate rocks

according to depositional texture. In: Classification of

Carbonate rocks – A Symposium (Ed W. E. Ham), Am.

Assoc. Petrol. Geol. Mem., 1, 108–121.Eberli, G.P. (2013) The uncertainties involved in extracting

amplitude and frequency of orbitally driven sea-level

fluctuations from shallow-water carbonate cycles.

Sedimentology, 60, 64–84.Eberli, G.P. and Ginsburg, R.N. (1987) Role of Cenozoic

progradation in evolution of Great Bahama Bank. AAPG

Bull., 71, 552–552.Enos, P. (1974) Surface sediment facies of the Florida-

Bahamas Plateau. GSA Map Series, MC-5, 4 edn.

Gonzalez, R. and Eberli, G.P. (1997) Sediment transport and

bedforms in a carbonate tidal inlet; Lee Stocking Island,

Exumas, Bahamas. Sedimentology, 44, 1015–1030.Hardie, L.A. and Garrett, P. (1977) General environmental

setting. In: Sedimentation on the Modern Carbonate Tidal

Flats of Norwest Andros, Bahamas (Ed L. Hardie), pp. 12–49. John Hopkins University Press, Baltimore.

Harris, P.M. (1979) Facies Anatomy and Diagenesis of a

Bahamian Ooid Shoal. VII, Sedimenta. University of

Miami, Miami, FL, 163 pp.

Harris, P.M. (2010) Delineating and quantifying depositional

facies patterns in carbonate reservoirs – insight from

modern analogs. AAPG Bull., 94, 61–86.Harris, P.M., Ellis, J. and Purkis, S. (2010) Delineating and

quantifying depositional facies patterns of modern

carbonate sand deposits on Great Bahama Bank. SEPM

Short Course, 54, 1–51, appendix p. 1-31, and 2 DVDs.

Harris, P.M., Purkis, S.J. and Ellis, J. (2011) Analyzing

spatial patterns in modern carbonate sand bodies from

Great Bahama Bank. J. Sed. Res., 81, 185–206.Harris, P.M., Purkis, S. and Ellis, J. (2014) Evaluating water-

depth variation and mapping depositional facies on Great

Bahama Bank – a “flat-topped” isolated carbonate

platform. SEPM Short Course, 56, 1–44 and DVD.

Hine, A.C. (1977) Lily Bank, Bahamas: history of an active

oolite sand shoal. J. Sed. Res., 47, 1554–1581.Hine, A.C., Wilber, R.J. and Neumann, A.C. (1981)

Carbonate sand bodies along contrasting shallow bank

margins facing open seaways in northern Bahamas. AAPGBull., 65, 261–290.

Illing, L.V. (1954) Bahaman calcareous sands. AAPG Bull.,

38, 1–95.



Kaczmarek, S.E., Hicks, M.K., Fullmer, S.M., Steffen, K.L. andBachtel, S.L. (2010) Mapping facies distributions on modern

carbonate platforms through integration of multispectral

Landsat data, statistics-based unsupervised classifications,

and surface sediment data. AAPG Bull., 94, 1581–1606.Kelly, M. and Tuxen, M. (2009) Remote sensing support for

tidal vegetation research and management. In: Remote

Sensing and Geospatial Technologies for Coastal

Ecosystem Assessment and Management (Ed X. Yang), pp.

341–362. Springer Verlag, Berlin.Lagerloef, G.S.E., Mitchum, G., Lukas, R and Niiler, P.

(1999) Tropical Pacific near surface currents estimated

from altimeter, wind and drifter data. J. Geophys. Res.,

104, 23,313–23,326.Loreau, J.P. (1982) Sediments aragonitiques et leur genese.

Mem. Mus. Nat. Hist. Nat. Ser. C, Tome XLVII, 300.Lucia, F.J. (1995) Rock-fabric/petrophysical classification of

carbonate pore space for reservoir characterization. AAPGBull., 79, 1275–1300.

Lumpkin, R., Goni, G. and Dohan, K. (2009) Surface Current

Observations, in “State of the Climate in 2008”. Bull. Am.Meteor. Soc., 90, S1–S196.

Milliman, J.D., Freile, D., Steinen, R.P. and Wilber, J.(1993) Great Bahama Bank aragonitic muds: mostly

inorganically precipitated, mostly exported. J. Sed.Petrol., 63, 589–595.

Morse, J.W., Millero, F.J., Thurmond, V., Brown, E. and

Ostlund, H.G. (1984) The carbonate chemistry of Grand

Bahama Bank waters – after 18 years another look. J. Geo.Res. Oceans, 89, 3604–3614.

Morse, J.W., Gledhill, D.K. and Millero, F.J. (2003) CaCO3

precipitation kinetics in waters from the Great Bahama

Bank: implications for the relationship between bank

hydrochemistry and whitings. Geochim. Cosmochim. Acta,

67, 2819–2826.Newell, N.D., Imbrie, J., Purdy, E.G. and Thurber, D.L.

(1959) Organism communities and bottom facies,

Great Bahama Bank. Bull. Am. Mus. Nat. Hist., 117, 177–228.

Palmer, M.S. (1979) Holcene facies geometry of the leeward

bank margin, Tongue of the Ocean, Bahamas. MS Thesis,

University of Miami, Miami, FL, 199 pp.

Purdy, E.G. (1963a) Recent calcium carbonate facies of the

Great Bahama Bank. 1. Petrography and reaction groups.

J. Geol., 71, 334–355.Purdy, E.G. (1963b) Recent calcium carbonate facies of the

Great Bahama Bank. 2. Sedimentary facies. J. Geol., 71,472–497.

Purkis, S.J. (2005) A “reef-up” approach to classifying coral

habitats from IKONOS imagery. IEEE Trans. Geosci.

Remote Sens., 43, 1375–1390.Purkis, S.J. and Klemas, V. (2011) Remote Sensing and

Global Environmental Change. Wiley-Blackwell, Oxford,

368 pp.

Purkis, S.J. and Pasterkamp, R. (2004) Integrating in situ

reef-top reflectance spectra with Landsat TM imagery to

aid shallow-tropical benthic habitat mapping. Coral Reefs,

23, 5–20.Purkis, S.J., Harris, P.M. and Ellis, J. (2012) Patterns of

sedimentation in the contemporary Red Sea as an analog

for ancient carbonates in rift settings. J. Sed. Res., 8, 859–870.

Purkis, S.J., Kerr, J., Dempsey, A., Calhoun, A., Metsamaa,L., Riegl, B., Kourafalou, V., Buckner, A. and Renaud, P.(2014a) Large-scale carbonate platform development of Cay

Sal Bank, Bahamas, and implications for associated reef

geomorphology. Geomorphology, 222, 25–38.Purkis, S.J., Rowlands, G. and Kerr, J.M. (2014b)

Unravelling the influence of water depth and wave energy

on the facies diversity of shelf carbonates. Sedimentology,

doi:10.1111/sed.12110.

Rankey, E.C. (2002) Spatial patterns of sediment accumulation

on a Holocene carbonate tidal flat, northwest Andros Island,

Bahamas. J. Sed. Res., 72, 591–601.Rankey, E.C. (2004) On the interpretation of shallow shelf

carbonate facies and habitats: how much does water depth

matter? J. Sed. Res., 74, 2–6.Rankey, E.C. and Doolittle, D.F. (2012) Geomorphology of

carbonate platform-marginal uppermost slopes: insights

from a Holocene analogue, Little Bahama Bank, Bahamas.

Sedimentology, 59, 2146–2171.Rankey, E.C. and Reeder, S.L. (2011) Holocene oolitic

marine sand complexes of the Bahamas. J. Sed. Res., 81,97–117.

Rankey, E.C. and Reeder, S.L. (2012) Tidal sands of the

Bahamian archipelago. In: Principles of TidalSedimentology (Eds R.A. Davis and R.W. Dalrymple), pp.

537–565. Springer-Verlag, Berlin.Rankey, E.C., Riegl, B. and Steffen, K. (2006) Form, function

and feedbacks in a tidally dominated ooid shoal, Bahamas.

Sedimentology, 53, 1191–1210.Rankey, E.C., Reeder, S.L., Ritter, S. and Harris, P.M. (2008)

Morphology and dynamics of carbonate tidal sand ridges:

Schooner Cays, Bahamas. AAPG Bull. InternationalConference Abstracts Volume, p. 32. (posted in 2008 on

AAPG Search and Discovery as Article #50134)

Reeder, S.L. and Rankey, E.C. (2008) Interactions between

tidal flows and ooid shoals, northern Bahamas. J. Sed.

Res., 78, 175–186.Reeder, S.L. and Rankey, E.C. (2009) Controls on

morphology and sedimentology of carbonate tidal deltas,

Abacos, Bahamas. Mar. Geol., 267, 141–155.Reid, R.P. and MacIntyre, I.G. (2000) Microboring versus

recrystallization: further insight into the micritization

process. J. Sed. Res., 70, 24–28.Reijmer, J.J.G., Swart, P.K., Bauch, T., Otto, R., Reunig, L.,

Roth, S. and Zechel, S. (2009) A re-evaluation of facies on

Great Bahama Bank I: new facies maps of western Great

Bahama Bank. In: Perspectives in Carbonate Geology: A

Tribute to the Career of Robert Nathan Ginsburg (Eds P.K.

Swart, G.P. Eberli and J.A. McKenzie), IAS Spec. Publ., 41,29–46.

Robbins, L.L. (1992) Biochemical and ultrastructural

evidence for the origin of whitings: a biological induced

calcium carbonate precipitation mechanism. Geology, 20,464–468.

Robbins, L.L., Tao, Y. and Evans, C.A. (1997) Temporal and

spatial distribution of whitings on Great Bahama Bank and

a new lime mud budget. Geology, 25, 947–950.Roth, S. and Reijmer, J.J.G. (2004) Holocene Atlantic climate

variations deduced from carbonate periplatform sediments

(leeward margin, Great Bahama Bank). Paleoceanography,

19, PA1003, doi:10.1029/2003PA000885.

Roth, S. and Reijmer, J.J.G. (2005) Holocene millennial to

centennial carbonate cyclicity recorded in slope sediments

of the Great Bahama Bank and its climatic implications.

Sedimentology, 52, 161–181.Schlager, W. and Purkis, S.J. (2013) Bucket structure in

carbonate accumulations of the Maldive, Chagos and

Laccadive archipelagos. Int. J. Earth Sci., 102, 2225–2238.



Shannon, C.E. (1948) A mathematical theory of

communication. Bell Syst. Tech. J., 27, 379–423.Shinn, E.A., Steinen, R.P., Lidz, B.H. and Swart, P.K. (1989)

Whitings, a sedimentologic dilemma. J. Sed. Pet., 59, 147–161.

Sparks, A.G. and Rankey, E.C. (2013) Relations between

geomorphic form and sedimentologic-stratigraphic

variability: Holocene ooid sand shoal, Lily Bank, Bahamas.

AAPG Bull., 97, 61–85.Steinen, R.P., Swart, P.K., Shinn, E.A. and Lidz, B. (1988)

Bahamian lime mud: the algae didn’t do it. Geol. Soc. Am.

Abstracts Program, 20, A209.

Swart, P.K., Reijmer, J.J. and Otto, R. (2009) A reevaluation

of facies on Great Bahama Bank II: Variations in the

d13C, d18O and mineralogy of surface sediments. In:

Perspectives in Carbonate Geology: A Tribute to the Career

of Robert Nathan Ginsburg (Eds P.K. Swart, G.P. Eberli

and J.A. McKenzie), IAS Spec. Publ., 41, 47–60.Swart, P., Oehlert, A.M., Mackenzie, G.J., Eberli, G.P. and

Reijmer, J.J. (2014) The Fertilization of the Bahamas by

Saharan Dust: a trigger for carbonate precipitation?

Geology, 42, 671–674.Traverse, A. and Ginsburg, R.N. (1966) Palynology of the

surface sediments of Great Bahama Bank, as related to water

movement and sedimentation. Mar. Geol., 4, 417–459.

Wang, L., Sousa, W.P. and Gong, P. (2004) Integration of

object-based and pixel-based classification for mapping

mangroves with IKONOS imagery. Int. J. Remote Sens., 25,5655–5668.

Wilber, R.J., Milliman, J.D. and Halley, R.B. (1990)

Accumulation of bank-top sediment on the western slope

of the Great Bahama Bank: rapid progradation of a

carbonate megabank. Geology, 18, 970–974.Wilson, P.A. and Roberts, H.H. (1992) Carbonate

periplatform sedimentation by density flows: a mechanism

for rapid off-bank transport and vertical flux of shoal

water fines. Geology, 20, 713–716.Wilson, P.A. and Roberts, H.H. (1995) Density cascading:

off-shelf sediment transport, evidence and implications,

Bahama Banks. J. Sed. Res., A65, 45–56.Yates, K.K. and Robbins, L.L. (1998) Production of carbonate

sediments by a unicellular green alga. Am. Mineral., 83,1503–1509.

Manuscript received 9 March 2014; revision accepted8 August 2014



mapping bathymetry and depositional facies on...

Documents