Geochemistry of meteoric diagenesis in carbonate islands of the northern Bahamas: 2. Geochemical modelling and budgeting of diagenesis

HYDROLOGICAL PROCESSESHydrol. Process. 21, 967–982 (2007)Published online 20 February 2007 in Wiley InterScience(www.interscience.wiley.com) DOI: 10.1002/hyp.6533

Geochemistry of meteoric diagenesis in carbonate islands ofthe northern Bahamas: 2. Geochemical modelling and

budgeting of diagenesis

Fiona F. Whitaker1* and Peter L. Smart2

1 Department of Earth Sciences, University of Bristol, Wills Memorial Building, Queens Road, Bristol BS8 1RJ, UK2 School of Geographical Sciences, University of Bristol, University Road, Bristol BS8 1SS, UK

Abstract:

Geochemical characterization and numerical modelling of surface water and ground water, combined with hydrologicalobservations, provide quantitative estimates of meteoric diagenesis in Pleistocene carbonates of the northern Bahamas.

Meteoric waters equilibrate with aragonite, but water- rather than mineral-controlled reactions dominate. Dissolutionallowering of the undifferentiated bedrock surface is an order of magnitude slower than that within soil-filled topographichollows, generating small-scale relief at a rate of 65–140 mm ka�1 and a distinctive pocketed topography. Oxidation oforganic matter within the subsoil and vadose zones generates an average PCO2 of 4Ð0 ð 10�3 atm, which drives dissolutionduring vadose percolation and/or at the water table. However, these dissolution processes together account for <60% of theaverage rock-derived calcium in groundwaters pumped from the freshwater lens. The additional calcium may derive fromoxidation of organic carbon within the lens, accounting for the high PCO2 of the lens waters. Mixing between meteoric watersof differing chemistry is diagenetically insignificant, but evapotranspiration from the shallow water table is an important drivefor subsurface cementation.

Porosity generation in the shallow vadose zone averages 1Ð6–3Ð2% ka�1. Phreatic meteoric diagenesis is focused near thewater table, where dissolution generates porosity at 1Ð4–2Ð8% ka�1. Maximum dissolution rates, however, are similar to thoseof evaporation-driven precipitation, which occludes porosity of 4Ð0 š 0Ð6% ka�1. This drives porosity inversion, from primaryinterparticle to secondary mouldic, vug and channel porosity. In the deeper freshwater lens, oxidation of residual organiccarbon and reoxidation of reduced sulphur species from deeper anaerobic oxidation of organic carbon may generate porosityup to 0Ð06% ka�1.

Meteoric diagenesis relies critically on hydrological routing and vadose thickness (controlled by sea level), as well as thegeochemical processes active. A thin vadose zone permits direct evaporation from the water table and drives precipitation ofmeteoric phreatic cements even where mineral stabilization is complete. Copyright 2007 John Wiley & Sons, Ltd.

KEY WORDS diagenesis; Bahamas; dissolution; meteoric; carbonate; geochemical modelling; eogenetic karst; rates of diagenesis

Received 9 August 2005; Accepted 23 June 2006

INTRODUCTION

Early meteoric diagenesis is an important control on boththe porosity and permeability of young carbonate rocks,and their later diagenesis and compaction (Budd, 2001;Moore, 2001). In order to predict the distribution andextent of early diagenesis in ancient carbonates, we needto understand the nature of the processes, their controlsand the rates at which they operate in present-day car-bonate settings. Quantitative estimates of the rates ofmeteoric diagenesis are useful for interpretation of thesedimentary record, but are essential for forward mod-elling of the diagenesis of carbonate platforms (Whitakeret al., 1997). The traditional geological approach to deter-mination of the processes of carbonate diagenesis is toexamine the diagenetic products preserved in the rockrecord. Where sequences are datable this gives a goodindication of long-term rates of diagenetic change. For

* Correspondence to: Fiona F. Whitaker, Department of Earth Sciences,University of Bristol, Wills Memorial Building, Queens Road, BristolBS8 1RJ, UK. E-mail: [email protected]

instance, the classic study of Land et al. (1967) identifiedthe sequence and rates for diagenetic alteration of min-eralogically immature carbonates. Comparative studies,for example in Barbados (Harrison, 1975), have demon-strated that rates of mineral stabilization are dependenton climate.

An alternative approach is to determine the present-day flux of dissolved carbonate from the hydrology andgeochemistry of surface and groundwater. Aqueous geo-chemistry is a relatively sensitive indicator of diagenesis,e.g. generation of porosity at a rate of 0.1% ka�1 (equiv-alent to 10�6 m3 porosity per m3 of rock year�1) with agroundwater flux of 0.3 m3 year�1 would give an increasein dissolved calcium of 3.6 mg l�1, which is readily mea-sured using modern analytical methods. This approachhas the further advantage of enabling the large-scale dis-tribution of processes and the rates of diagenesis to beinvestigated. The studies of Budd (1988) and McClainet al. (1992) on rates and processes of mineral stabiliza-tion in Holocene grainstones are excellent examples ofthis approach. However, the seminal work of Plummer

Copyright 2007 John Wiley & Sons, Ltd.

968 F. F. WHITAKER AND P. L. SMART

et al. (1976) on Bermuda remains one of the few majorintegrative studies of meteoric water diagenesis of pre-Holocene carbonates.

In a companion paper, Whitaker and Smart (2007)show that field sampling of surface waters and groundwaters can provide important new insights into the geo-chemical processes occurring in the Plio-Pleistocene car-bonates of the northern Bahamas. However, this studyincluded no water samples from the thin vadose zone.Furthermore, although critical processes (such as the con-tribution of the oxidation of dissolved organic matterto dissolution within the meteoric freshwater lens) weredemonstrated to be plausible, the quantitative significanceof these processes was not investigated. Geochemical for-ward modelling provides a partial solution to these prob-lems, enabling quantitative testing of specific processesand the exploration of different evolutionary pathwayson the distribution of dissolution and reprecipitation ofcarbonate (e.g. Plummer, 1977).

In this second paper considering meteoric diagenesisin the northern Bahamas, we quantitatively examinethe relative significance of different processes drivingdiagenesis to understand porosity evolution in youngcarbonates better. We use geochemical modelling toexplore the evolution of poorly known parts of thehydrochemical system to understand and constrain theprocesses of dissolution and precipitation of carbonatewithin the Pleistocene carbonates. The aqueous speciationmodel PHREEQC (Parkhurst, 1995) was used for forwardmodelling, and simulations were constrained using theanalytical data described in Whitaker and Smart (2007).The results from this modelling study are then combinedwith those from the field study to generate a distributedbudget of carbonate dissolution and precipitation for thePleistocene carbonates of the northern Bahamas that arecompared with results from other hydrochemical studiesof carbonate islands.

STUDY AREA

The study draws on data from three of the largest islandsof the northern Bahamas: North and South Andros Islandson the Great Bahama Bank, and Grand Bahama on thesouthern edge of the Little Bahama Bank (Figure 1). Onthese islands the carbonates are dominated by poorlystratified non-skeletal shallow-water limestones of thelate Pliocene–Pleistocene Lucayan Formation, withinwhich there are frequent subaerial exposure horizons(Beach and Ginsburg, 1980). This formation also includescomplex sequences of low linear and arcuate aeolian andbeach dune ridges, which dominate present-day relief(Garrett and Gould, 1984). Terrain is determined largelyby subtle differences in elevation, grading from dryrocklands, through zones of marsh, lagoon and swamp,to shallow lakes. Subaerial Holocene deposits are onlylocally present as beach ridges and tidal flats.

The northern Bahamas has a subtropical marine cli-mate, with persistent trade winds, a warmer rainy season

Figure 1. Map of the northern Bahamas showing location of namedislands and regional variation in rainfall (contours in mm year�1)

from May to October and a cooler drier season fromNovember to April. There is a broad north–south climaticgradient (Whitaker and Smart, 1997a; Figure 1), the wet-ter northern Bahamas having almost twice the annualrainfall of the more arid southern islands. Within thestudy area, mean annual rainfall reduces from 1496 mmin Freeport (Grand Bahama) to 1175 mm at Kemps Bay(South Andros). Temperatures are semi-tropical with lim-ited diurnal and seasonal fluctuations, from 24–33 °Cin the summer to 17–27 °C in the winter (maximumand minimum average daily temperatures for Augustand January). Potential evapotranspiration has been esti-mated from Bahamas Meteorological Office data for NewProvidence (Whitaker, 1992) as 1610 mm year�1 (Pen-man) and 1581 mm year�1 (corrected open pan). How-ever, rates of actual evapotranspiration are likely sub-stantially lower. Little et al. (1977) calculated a meaneffective recharge of 375 mm year�1 for Grand Bahamaand 300 mm year�1 for Cat Island and Eleuthera (clima-tologically comparable to South Andros), giving actualevapotranspiration rates of 1121 mm year�1 and 875 mmyear�1 for Grand Bahama and South Andros respectively(¾75% of total rainfall).

A proportion of this evaporative demand is satisfiedfrom interception retained on vegetative surfaces. GrandBahama and North Andros host extensive pine forestswith a sparse understorey of palm, with an increasein coppice woodland on the drier islands, includingSouth Andros (Campbell, 1978). Surface retention insubtropical pine forests is estimated as 10–20% oftotal rainfall (Balek, 1983; Crockford and Richardson,1990). This gives estimates of total interception (Table I)that are comparable to those derived from the productof retention (1–3 mm per event; Falkland and Brunel,1993) and the number of events (approximated by thenumber of rain days: 122 on Grand Bahama and 76 onSouth Andros). The residual evapotranspirative deficit,

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 967–982 (2007)DOI: 10.1002/hyp

GEOCHEMISTRY OF METEORIC DIAGENESIS IN CARBONATE ISLANDS: 2 969

Table I. Hydrodynamics of freshwater recharge and residence time in selected islands of the northern Bahamas. Data from Littleet al. (1977), Cant and Weech (1986), and unpublished data from the Bahamas Meteorological Office

GrandBahama

North Andros South Andros

Mean rainfall (m year�1) 1Ð496 1Ð340 1Ð175Estimated interceptiona (m year�1) 0Ð224 0Ð201 0Ð176Estimated evapotranspiration (m year�1) 1Ð121 1Ð001 0Ð875Effective recharge (m year�1) 0Ð375 0Ð339 0Ð300Area underlain by fresh water (m2) 5Ð92 ð 108 1Ð37 ð 109 8Ð18 ð 108

Fresh water lens volumeb (m3) 1Ð53 ð 109 4Ð31 ð 109 1Ð50 ð 109

Maximum fresh water thickness (m) 20Ð6 34Ð0 NDAverage fresh water thickness (m) 8Ð60 10Ð5 6Ð09Average fresh water residence time (year) 6Ð90 9Ð28 6Ð09

a Assuming interception is 15% of mean annual rainfall.b Assuming porosity is 30%.

equivalent to 60% of total rainfall, is satisfied from thesubsurface (Whitaker, 1992).

The high secondary permeability of the karstifiedlimestones gives rapid infiltration of surface runoff viavertical fissures, fractures and root channels. However,slower diffuse percolation via intergranular voids inthe high-porosity carbonates can also occur. Limitedfield observations suggest that rapid percolation duringheavy rainfall is predominant, being facilitated by theabundance of surface dissolution features that focus flow(Mylroie and Carew, 1995) and the development ofmicrite crusts on the surface of the carbonates (Rossinskiand Wanless, 1992). The depth of the vadose zoneis controlled largely by topography, and over largeareas remains less than 1 m, although it may exceed5 m beneath aeolianite ridges. The average depth ofthe vadose zone measured in observation boreholesdistributed across North Andros during the rainy seasonby Little et al. (1973) is only 0Ð8 š 0Ð7 m �n D 56�.Although we have no data on vadose residence times,there appears to be limited long-term vadose storage.The response of the water table to storm recharge events(defined as events with total rainfall ½20 mm) is rapid,with some 45–75% of rainfall reaching the water tablewithin 12 h of rainfall (Figure 2; data from Little et al.(1973, 1975)). This response appears to be independentof both the total rainfall and the antecedent moistureconditions. In the high-permeability limestones (Whitakerand Smart, 1997a), water table elevations return to pre-storm levels within 24–48 h, probably due largely toadjustments in the position of the base of the freshwaterlens.

In the saturated zone, meteoric groundwater occurs asGhyben–Herzberg lenses, overlying more dense salinegroundwaters. The nature and size of the freshwa-ter lens is determined by island size and shape, cli-mate and aquifer characteristics (Cant and Weech, 1986;Whitaker and Smart, 1997b). Throughout the study areathe actual freshwater lens is thinner than the theoreticalGhyben–Herzberg lens (Cant and Weech, 1986; Vacherand Bengtson, 1989) because the hydraulic conductivity

Figure 2. Response of the water table to rainfall (at time zero) measuredin a borehole on Grand Bahama, corrected for tidal and barometricpressure variations. Relative water level change is expressed as apercentage of maximum rainfall-induced water level change to facilitatecomparison of different magnitude rainfall events. Data from Little

et al. (1975)

of the upper Lucayan Limestone aquifer increases sub-stantially with depth (Whitaker and Smart, 1997a). OnNorth Andros the freshwater lens reaches a maximumthickness of 34 m, limited by the distance between tidalcreeks, which function as estuaries discharging fresh andbrackish groundwater (Whitaker and Smart, 1997b). Onthe Little Bahama Bank, where rates of subsidence arerelatively low, the base of the freshwater lens is trun-cated at 21 m at the boundary between the LucayanFormation and the older, more conductive limestonesbelow (Cant and Weech, 1986). Groundwater flow ispredominantly horizontal, with typical freshwater lensradii being 10–12 km on North Andros and 3–4 km onGrand Bahama. Using estimates of freshwater lens vol-ume from Cant and Weech (1986) and effective rechargetotals above, and assuming 30% average porosity, we canestimate the average residence time of water within the

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 967–982 (2007)DOI: 10.1002/hyp

970 F. F. WHITAKER AND P. L. SMART

freshwater lens. This ranges from 6 years and 7 years onSouth Andros and Grand Bahama respectively to over9 years on North Andros (Table I).

METHODS

Geochemical modelling used data for samples of meteoricwater collected from a range of sites and geochemicalenvironments on North and South Andros and GrandBahama (data in Whitaker and Smart (2007); Figure 1).Rainwater samples represent the input to the system, andthe resulting runoff from a variety of bare and vegetatedbedrock surfaces was sampled during and immediatelyafter storm events. Water retained at the bedrock surfacefor periods from a few hours to a few days was collectedfrom ephemeral surface pools and soil-filled solutionhollows, including metre-diameter dissolution pits locallytermed ‘banana holes’ (Smart and Whitaker, 1989). Themeteoric freshwater lens was sampled using boreholespumping and static boreholes, shallow hand-dug wellsand a variety of types of flooded cave (see Whitaker andSmart (1997b, 1998)).

We use chloride as a conservative tracer to differenti-ate between conservative processes and those involvingreactions between gas, mineral and organic phases. Theconcentrations of calcium, strontium, magnesium and sul-phate in excess (XS) of those predicted from simpledilution of local Tongue of the Ocean (TOTO) seawaterwere calculated, e.g.:

CaXS D CaSAMPLE � �CaTOTO ð ClSAMPLE/ClTOTO��1�

Root-mean-squared (RMS) errors were calculated bycombining uncertainties in estimates of sample andTOTO concentrations. The combined error for an averagepumping borehole water amounts to š3Ð3 mg l�1 CaXS,š0Ð44 mg l�1 SrXS, š7Ð8 mg l�1 MgXS and š14 mg l�1

SO4XS. Because the majority of the error derives fromuncertainty in TOTO concentrations, combined errors donot vary greatly between different meteoric waters.

The aqueous speciation model PHREEQC (Parkhurst,1995) was used to calculate the distribution and activitiesof all aqueous species, the partial pressure of CO2,ion activity products (IAP), and saturation states (logIAP/K) of aragonite (SIA) and calcite (SIC) using thethermodynamic equilibrium constants of Plummer andBusenberg (1982). Uncertainties in saturation indicesand PCO2 values are typically 0Ð036 log units and0Ð28 ð 10�3 atm respectively, with >95% of uncertaintyderiving from measurement of pH (š0Ð02 pH units).

In simulations, waters were equilibrated with respectto aragonite rather than the dominant mineral calcite.Whitaker and Smart (2007) demonstrate that the meteoricwaters tend to dissolve carbonate to reach equilibriumwith respect to aragonite, the most soluble phase present,even in Pleistocene rocks where only minor amounts ofaragonite remain (generally �10%; Beach, 1995). Vadosewaters are also allowed to react to equilibrium witha specified partial pressure of CO2 in ground air, the

PCO2 of which is constrained by field measurements ofWhitaker and Smart (2007). The amount of aragoniteprecipitated or dissolved to equilibrium is calculated.

Previous studies simulating the effect of groundwa-ter mixing on carbonate dissolution have focused almostexclusively on the fresh-water–salt-water mixing zone(e.g. Sanford and Konikow, 1989). Here, we examinethe effect of mixing within the meteoric system betweenrapid and diffuse percolation waters and between watersat top and at depth within the freshwater lens. Two solu-tions of defined composition are mixed in specified pro-portions and the mixture is then equilibrated with arag-onite to investigate the mass of the mineral precipitatedor dissolved to reach aragonite equilibrium.

Evaporation is simulated by removing water from thechemical system via negative titration with pure water(water is specified as an irreversible reactant with a neg-ative reaction coefficient). For a representative freshwaterlens top sample, the amount of water to be removedis specified based on evapotranspirative deficit satis-fied from the subsurface (equivalent to 60% of thetotal rainfall; Whitaker and Smart, 2007). The num-ber of moles of all elements is then multiplied, effec-tively increasing the mass of the aqueous phase whilstmaintaining the concentration. We assume that evapo-ration and evapotranspiration have the same effect andthat evapotranspiration has no effect on the ion ratios(although this is poorly constrained and may not bethe case). After evaporation we calculate the saturationof the solution with respect to aragonite and CO2, andthe amount of aragonite that would precipitate to reachequilibrium.

Oxidation of organic carbon is simulated by adding adefined amount of carbon irreversibly to a representativefreshwater lens sample from a pumping borehole. Theamount of carbon reacted is constrained by measurementsof dissolved organic carbon (DOC) within groundwa-ter (Whitaker and Smart, 2007). Aragonite is allowed todissolve to equilibrium if the solution becomes undersat-urated with aragonite.

Uncertainty in values that are derived from more thanone parameter estimate is handled using the RMS error,combining fractional errors for different componentsof the estimate, assuming that all errors are randomlydistributed.

GEOCHEMICAL MODELLING

Geochemical evolution of percolation waters in thevadose zone

Dissolution by waters percolating through the vadosezone is controlled by the hydrological routing and resi-dence time within the vadose zone and the PCO2 of groundair. Figure 3 is a schematic summary of the hydrologicalsystem, indicating possible routes for vadose percolationand their relative volumetric importance. Table II sum-marizes the results of geochemical modelling of differentvadose percolation scenarios.

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 967–982 (2007)DOI: 10.1002/hyp

GEOCHEMISTRY OF METEORIC DIAGENESIS IN CARBONATE ISLANDS: 2 971

Figure 3. Schematic diagram illustrating hydrology of the meteoricsystem for the islands of the Northern Bahamas. Figures in metres peryear related to water budget for North Andros. Percentages indicate

relative volumetric importance of surface and vadose routing

Waters running off from bare and vegetated bedrocksurfaces to infiltrate the vadose zone are equilibratedwith atmospheric PCO2 , but are undersaturated withrespect to both aragonite and calcite (SIA: �1Ð37 š 0Ð47;SIC: �1Ð23 š 0Ð47; Whitaker and Smart, 2007). There-fore, they have some limited potential for subsurface

dissolution. Vadose dissolution Cavadose under theseconditions could range from zero (where infiltration viahigh-flow routes is so rapid that it completely inhibitsrock–water interaction; Table II, scenario A.1) to a max-imum (where rapid percolation waters move more slowlyand equilibrate with respect to aragonite; Table II, sce-nario A.2). In the case of the latter, the geochemicalmodelling suggests that, on average, an additional 15 mgl�1 calcium would dissolve, based on average surfacerunoff driven to equilibrium with aragonite (Table II, sce-nario A.2).

In Part 1 (Whitaker and Smart, 2007) we arguedthat, despite the presence of open fissures within thevadose zone, the ground air has an elevated PCO2 . Thisis generated and maintained by downward diffusion ofCO2 from the surface soil, CO2 production by treeroots that penetrate the rock to depths of several metres,bacterial oxidation of organic matter washed down fromthe surface, and upwards diffusion of CO2 from thefreshwater lens (Whitaker and Smart, 2007). We suggestthat the minimum PCO2 of ground air is determined by thePCO2 of the soil atmosphere, and can be estimated fromwater samples from soil-filled pockets (banana holes)and from the degassed waters in the upper part of thefreshwater lens. Here, for modelling we employ a valueof �4Ð0 š 1Ð1� ð 10�3 atm based on our field sampling.Significant additional carbonate dissolution may occur ifgroundwaters percolating through the vadose zone havea sufficiently long residence time to equilibrate bothwith this elevated ground-air PCO2 and the aragonitepresent (Table II, scenario A.3). This will most likelyoccur for recharge via diffuse intergranular percolation.

Table II. Geochemical modelling of vadose percolation waters showing the effect of hydrological routing on carbonate chemistry

Scenario Hydrological routing and residence time controlmixture of percolation waters

Average composition of percolation watersreaching the water table

Comparison with watersfully equilibrated at

the water table

Rapid percolation (%) Diffusepercolationc

(%)

Banana-holepercolationd

(%)

PCO2

(10�3 atm)SIA

(log IAP/K)CaXS

(mg l�1)Cavadose

e

(mg l�1)CaWtequil

(mg l�1)Cavadose :CaWTequil

f

Noequilibrationa

Withequilibrationb

Scenarios A: end-membersA.1 100 0 0 0 0Ð3 �1Ð33 7Ð7 0 46 90 : 100A.2 0 100 0 0 0Ð3 0Ð00 22 15 32 32 : 68A.3 0 0 100 0 4Ð0 0Ð00 54 47 0 100 : 0A.4 0 0 0 100 4Ð5 0Ð02 51 0 �3 —

Scenarios B: mixing of rapid (direct and banana hole) and diffuse percolationB.1 55 0 25 20 2Ð0 �0Ð46 28 12 26 32 : 68B.2 0 55 25 20 1Ð8 �0Ð13 36 20 18 53 : 47B.3 25 0 55 20 3Ð1 �0Ð17 41 24 13 65 : 35B.4 0 25 55 20 3Ð0 �0Ð06 41 24 13 65 : 35

a With respect to CO2 or aragonite.b At atmospheric PCO2 and aragonite equilibrium.c At ground-air PCO2 and aragonite equilibrium.d Rapid percolation from banana holes at soil-water PCO2 and aragonite equilibrium (measured banana-hole waters).e Calcium derived from vadose-zone dissolution.f Calcium derived from dissolution at the water table due to equilibration of vadose percolation water with ground-air PCO2 and aragonite.

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 967–982 (2007)DOI: 10.1002/hyp

972 F. F. WHITAKER AND P. L. SMART

Geochemical modelling suggests an additional 47 mg l�1

calcium would be dissolved for recharge of this type.Water can also percolate rapidly via metre-diameter

dissolution pits locally termed ‘banana holes’ (Smartand Whitaker, 1989; note that our usage of this termdiffers from that of Harris et al. (1995)). Banana holesare generally soil filled, but characteristically are drainedby fissures that permit rapid percolation to the water table.Sampled waters from banana holes have an elevated PCO2

similar to that predicted for ground air (�4Ð5 š 2Ð8� ð10�3 atm). They are close to equilibrium with respectto aragonite, having dissolved an average of 51 mg l�1

calcium. If recharge causes piston displacement, then thishigh-calcium soil water would pass through the vadosezone without further geochemical evolution, and noadditional bedrock dissolution (Table II, scenario A.4).

The relative importance of these four end-members interms of total recharge is relatively poorly constrainedby hydrological observations. The evidence from theresponse of the water table to storm events is that rapidpercolation predominates, although the degree to whichthis equilibrates with aragonite is not well known. There-fore, we simulate two scenarios (Table II, scenarios B):one with predominately (75%) rapid recharge and onewith more diffuse percolation (45% rapid recharge). Ofthe rapid recharge component, we estimate that a sig-nificant fraction will be routed via fissures at the baseof soil-filled hollows such as banana holes. We havedefined the area of banana holes, together with contrib-utory catchments for surface runoff, from topographicsurveys at several sites on North Andros representativeof different terrain types. This suggests that rapid per-colation includes 20% banana-hole water. This water isalready equilibrated with aragonite at high soil PCO2 andhas a high calcium excess. Mixing of diffuse vadosepercolation, and direct and banana-hole rapid percola-tion (Table II, scenarios B) generates a range of possibleestimates of the increase in CaXS due to vadose zone dis-solution of 12–24 mg l�1. Estimates are more sensitiveto uncertainties in the ratio of rapid : diffuse percolation(Table II, scenario B.1 versus B.3 and scenario B.2 versusB.4) then to the extent to which rapid percolation watersequilibrate with atmospheric CO2 (Table II, scenario B.1versus B.2 and scenario B.3 versus B.4).

All simulated percolation waters are aragonite under-saturated (Table II, scenarios B), with estimates of arag-onite saturation index ranging from �0Ð06 to �0Ð46.Modelling indicates that these waters have the potentialto dissolve a further 0Ð7–2Ð2 mg l�1 calcium to reachequilibrium with respect to aragonite. More significantis that the estimated PCO2 of all simulated mixtures ofvadose percolation waters (�1Ð8–3Ð1� ð 10�3 atm) is sig-nificantly less than that of the lens-top waters, which areequilibrated with PCO2 of ground air (4Ð0 ð 10�3 atm).When equilibrated with respect to aragonite and a PCO2

of 4Ð0 ð 10�3 atm, as must occur at the water table, thepercolation waters can dissolve a further 13–26 mg l�1

calcium to reach a total calcium of 54 mg l�1 for all sim-ulated mixtures (sum of CaXS and CaWTequil; Table II,scenarios B).

Hydrological routing of percolation waters within thevadose zone thus has no effect on the total amount of dis-solution driven by ground-air CO2, but is very significantin controlling the relative proportions of vadose : lens-topdissolution. Dissolution at the water table due to equili-bration of percolation waters with respect to ground-airCO2 and aragonite dominates when percolation is largelyrapid with no equilibration, accounting for up to 68%of total dissolution in the modelled system (Table II,scenario B.1). In contrast, when diffuse percolation is sig-nificant, most (65%) of the additional calcium is derivedfrom dissolution during vadose percolation (Table II, sce-narios B.3 and B.4).

Mixing

Groundwater mixing can generate the potential fordissolution or precipitation due to the non-linearity ofmineral solubility as a function of salinity, PCO2 , tem-perature and ionic strength (Bogli, 1964; Runnels, 1969).Thrailkill (1968) proposed that mixing of rapid rechargewaters with phreatic groundwater may cause undersatu-ration and, thus, drive bedrock dissolution. Mylroie andCarew (1990) invoked this mechanism to drive watertable dissolution in their flank margin model for cavedevelopment in carbonate islands. We have simulatedmixing between the four end-member vadose percolationwaters described above, between percolation and lens-topwaters, and between shallow and deeper waters within thefreshwater lens. We consider the modelled effects of mix-ing on aragonite saturation (Figure 4a and c) and also onthe change in CaXS (compared with conservative mixingbetween end-members) for mixtures driven to aragoniteequilibrium (Figure 4b and d).

Mixing between the different components of vadosepercolation (Table II, scenarios A) generates mixtures(Table II, scenarios B) that are less undersaturated withrespect to aragonite than would be expected from conser-vative mixing. Although in all cases the resultant mixedwater will still dissolve aragonite, the amount of calciumdissolved when the mixed waters equilibrate with respectto aragonite is very slightly (up to 0Ð5 mg l�1) reducedcompared with conservative mixing. Mixing also occursbetween percolation waters, which range from undersatu-rated to aragonite equilibrium (Table II), and waters at thesurface of the freshwater lens that have a higher PCO2 buthave degassed and are, hence, supersaturated (Whitakerand Smart, 2007). Mixing of any combination of vadosepercolation waters with lens-top waters also generatesmixed waters that are less undersaturated with respect toaragonite than expected (Figure 4a). When these mixedwaters are equilibrated with aragonite, the CaXS is upto 2Ð1 mg l�1 less than expected from conservative mix-ing (Figure 4b). The geochemical modelling, therefore,demonstrates that, despite previous suggestions, mixingprocesses are not a major drive for carbonate dissolutionat the top of the meteoric freshwater lens.

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 967–982 (2007)DOI: 10.1002/hyp

GEOCHEMISTRY OF METEORIC DIAGENESIS IN CARBONATE ISLANDS: 2 973

Figure 4. Simulated effect of mixing between meteoric waters on aragonite saturation index (a and c) and change in CaXS for mixtures (comparedwith conservative mixing) when equilibrated with aragonite (b and d). (a) and (b) show the effect of mixing runoff waters evolved according toscenarios A.1 to A.4 (solid lines, A.4 is similar to A.3) and B.1 to B.4 (dashed lines, B.4 is similar to B.3), as detailed in Table II, with lens-top water.(c) and (d) show the effect of mixing shallow lens waters (solid line is mixing with degassed surface water, dashed line is mixing with non-degassedshallow lens end-member) with water from the base of the brackish lens in Elvenholm Blue Hole, South Andros. Triangles are water samples froma vertical profile through the brackish lens in Elvenholm; circles identify end-members. In (d) the triangles show change in CaXS for water samples

equilibrated with aragonite

Finally, mixing may occur within the freshwater lensas well as between lens waters and the underlying salinegroundwater. This is indicated by the increasing salin-ity with depth observed within boreholes and, moremarkedly, in caves where mixing is enhanced in theopen water column (Whitaker, 1992; Whitaker and Smart,1997b). We simulated mixing within Elvenholm, a ver-tically extensive fracture cave on South Andros with abrackish lens (Figure 4c and d). The more brackish end-member water is a sample from the base of the lensand is close to aragonite equilibrium; the fresher end-members are a non-degassed upper lens sample (3 mwater depth) and a sample from the surface of the lenswhere degassing gives a significantly lower PCO2 andhigher carbonate saturation. Mixing with the degassedlens-top sample results in mixtures with a lower arago-nite saturation index, equivalent to a CaXS up to 0Ð34 mgl�1 higher on equilibration with aragonite. Mixing withthe non-degassed upper lens sample leads to a smallerdegree of additional undersaturation and a CaXS onlymarginally higher than predicted from conservative mix-ing. However, the aragonite saturation index of waterssampled in a profile through the water column is consid-erably lower (more negative) than predicted by conser-vative mixing (Figure 4c; Whitaker, 1992). Equilibrationof these samples with aragonite could increase the CaXS

by up to 4Ð5 mg l�1, suggesting that significant dissolu-tional potential exists within the bulk of the lens, drivenby some processes(es) other than mixing.

Evapotranspiration

The rapid infiltration of surface runoff and the shal-low depth of the vadose zone mean that, in the northernBahamas, a significant proportion of total evapotranspira-tion occurs from the saturated zone. We have simulatedthe effect of evapotranspirative concentration of watersby a factor of 2Ð5 times, equivalent to (total evapotran-spiration less interception)/effective recharge (Table I),on the carbonate geochemistry of meteoric waters inthe saturated zone. Lens-top water equilibrated with aground-air PCO2 of 4 ð 10�3 atm has a CaXS of 54 mgl�1 and is initially at equilibrium with respect to arago-nite. Geochemical modelling indicates that evaporativeconcentration causes a high degree of supersaturationwith respect to aragonite (SIA: C0Ð62). This will driveprecipitation of calcite to regain aragonite equilibrium.Modelling indicates precipitation of some 37 mg calciumper litre of water evaporated (at ground-air PCO2 ), indi-cating that a substantial portion of the calcium dissolvedfrom the bedrock may be reprecipitated from watersdirectly affected by evapotranspiration. Depending uponthe locus of evapotranspiration, this process could resultin development of a cemented zone in the upper part ofthe freshwater lens, a groundwater caliche at the watertable, cementation of the vadose zone or a case-hardenedmicritized surface crust. Evaporation from the very shal-low part of the vadose zone, where ground-air PCO2 islower, could result in less precipitation, as the source

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 967–982 (2007)DOI: 10.1002/hyp

974 F. F. WHITAKER AND P. L. SMART

waters may have a lower calcium excess. Evapotranspi-ration is thus a major process controlling the distributionand net flux of carbonate in oceanic carbonate terrains,and it can drive precipitation of more than 10 times theamount of cements than potentially driven by mixing.

In situ oxidation of organic matter

Whitaker and Smart (2007) showed that pumpedgroundwater from the freshwater lens had a much higherPCO2 than the soil atmosphere (�16 š 8Ð3� ð 10�3 atmcompared with �7Ð4 š 3Ð7� ð 10�3 atm) and suggestedthat bacterial decomposition of organic matter within thelens is an important process. Most waters sampled fromthe freshwater lens are sub-oxic, indicating consumptionof oxygen by aerobic decomposition of organic mat-ter. However, the widespread sulphate depletion in thefreshwater lens suggests that organic matter oxidationby sulphate reduction must also occur within microen-vironments and/or at depth within the freshwater lensfollowing consumption of all available oxygen. We havemodelled the effect of the oxidation of organic carbonon carbonate dissolution for both aerobic oxidation andsulphate reduction.

We simulated the addition of organic carbon to per-colation waters after equilibration with ground-air CO2

and aragonite and evapotranspiration (Figure 5). In prac-tice, we believe that evaporative concentration and oxi-dation of organics are probably simultaneous (Whitakerand Smart, 2007). Under aerobic conditions, the oxidationof 17Ð3 mg l�1 DOC, equivalent to the average DOC ofpumping boreholes (Whitaker and Smart, 2007), reducesthe aragonite saturation index to �1Ð02. When this wateris then equilibrated with respect to aragonite, carbonatedissolution increases the calcium concentration by 38 mgl�1. The modelled PCO2 (18 ð 10�3 atm) and calcium(92 š 12 mg l�1) of the resulting water are very simi-lar to those actually observed in samples from pumpingboreholes (�16 š 8Ð3� ð 10�3 atm and 93 š 18 mg l�1

calcium), strongly suggesting that the oxidation processsimulated is responsible for the observed groundwatergeochemistry.

Where the rate of oxygen diffusion downward from thewater table is exceeded by that of consumption by aerobicoxidation of organic carbon, the system becomes anaer-obic and oxidation occurs by sulphate reduction. Thisprocess is known to occur in the freshwater lenses ofthe northern Bahamas because a significant proportionof water samples from the saturated zone show sulphatedepletion (Whitaker and Smart, 2007). This anaerobicprocess also enhances undersaturation by the partial dis-sociation of H2S, which lowers the pH and has a greaterimpact on carbonate saturation than that of increasedalkalinity (Morse and Mackenzie, 1990). However, sul-phate reduction is only about half as efficient per mole ofcarbon in driving carbonate dissolution as oxidation underaerobic conditions (Figure 5). Anaerobic oxidation of17Ð3 mg l�1 organic carbon reduces the aragonite satura-tion index to �0Ð64, increases PCO2 to 22 ð 10�3 atm and

Figure 5. Modelled aragonite saturation index (dashed lines) and resultingCaXS (solid lines) when waters equilibrated with respect to aragonitefollowing oxidation of organic carbon in an aerobic system and bysulphate reduction in an anaerobic system. Vertical dashed line is average

DOC for samples from pumping boreholes

decreases sulphate concentration by 58 mg l�1. Equilibra-tion of this water with aragonite generates an increase incalcium of 18 mg l�1 and a PCO2 of 11 ð 10�3 atm. Suchwaters are observed at depth within the brackish lensof bank-marginal fracture caves (Whitaker and Smart,1997c) at the distal end of the meteoric groundwatercirculation.

Reduced sulphur species, including gaseous H2S, aregenerated by anaerobic oxidation of organic carbon atdepth. H2S diffuses along the concentration gradientwhich results from oxidation at the interface with oxicwaters. The re-oxidation of reduced sulphur species mayfurther drive carbonate dissolution (Whitaker and Smart,2007). The complete reoxidation of reduced sulphurspecies generated in the anoxic simulation describedabove produces further acidity, in this case sufficientto dissolve an additional 20 mg l�1 calcium to reacharagonite equilibrium. This reoxidation process couldaccount for the sulphate enrichment observed in somesamples from the upper oxic part of the freshwater lensand, in particular, from the fracture caves (Whitaker andSmart, 2007).

In conclusion, oxidation of organic carbon is capableof driving dissolution of 38 mg l�1 calcium. These calcu-lations assume that the entire fraction of dissolved carbonis available for oxidation, but a significant proportion oforganic carbon will be refractory and, thus, probably notreadily available. However, our calculations exclude par-ticulate carbon, which can represent 10–15% of the totalorganic carbon in pumping boreholes (Whitaker, 1992).Furthermore, immediately after heavy recharge events,which likely wash surface-derived organic carbon intothe subsurface, organic carbon concentrations in lens-topwaters may be considerably elevated. Because a signif-icant fraction of this carbon input may subsequently beoxidized, our simulations based on a total DOC of waters

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 967–982 (2007)DOI: 10.1002/hyp

GEOCHEMISTRY OF METEORIC DIAGENESIS IN CARBONATE ISLANDS: 2 975

pumped from a range of depths within the freshwater lensmay underestimate the importance of these processes atshallow depth. The total dissolution driven by oxidationof organic carbon is independent of whether the fresh-water lens is aerobic or anaerobic (with sulphate reduc-tion followed by reoxidation of reduced sulphur species).However, the redox boundary between aerobic and reduc-ing groundwaters will be a focus for dissolution. This willbe superimposed on a general trend of reducing availabil-ity and reactivity of organic matter with depth below thewater table, favouring dissolution in the upper part of thefreshwater lens.

A DISTRIBUTED BUDGET FOR METEORICCARBONATE DIAGENESIS

By combining measurements of CaXS in water sampleswith results from geochemical simulations we can gen-erate a distributed budget for carbonate dissolution andprecipitation for the meteoric system in the northernBahamas. The likely contribution of different processesto the calcium budget is summarized in Figure 6 andTable III. Figures in bold in the first column of Table IIIare the mean and standard deviation for the measuredcalcium concentrations in water samples from differenthydrochemical environments in the northern Bahamas

(from Whitaker and Smart (2007)). Other figures arederived from geochemical modelling of the possible evo-lution pathways under a number of scenarios, constrainedby hydrological observations. The second column indi-cates incremental change in CaXS attributable to partic-ular processes. These concentrations are corrected forthe contribution of sea salt using chloride as a conser-vative tracer (Equation (1)), and are thus representativeof changes due to dissolution of the carbonate bedrock.

Direct sampling of rain, runoff and soil waters indi-cates that dissolution at the undifferentiated bedrock sur-face contributes 6Ð6 š 2Ð1 mg l�1 calcium, only 13% ofthe 50 š 20 mg l�1 increase in calcium for sub-soil sur-face dissolution. These figures combine to give a totalCaXS for surface dissolution of 15 š 7Ð7 mg l�1 calcium,assuming that 20% of the water infiltrates via soil-filledpockets, such as banana holes (Table III). We have notconsidered near-surface reprecipitation of carbonate byevaporation of surface detention, as we are not able toestimate the volumetric extent of this process.

The geochemical modelling described above sug-gests that vadose dissolution is strongly dependent onhydrological routing. Calcium concentration in vadosewaters could increase by between 12 mg l�1 (75% rapidrecharge, with predominantly no subsurface carbonateequilibration) and 24 mg l�1 (55% diffuse recharge, with

Figure 6. Calcium budget for the meteoric system on North Andros, derived from geochemical sampling (diamonds) and numerical modelling(circles), describing the rate and distribution of dissolution and precipitation driven by different processes (see Table IV). Error bars are plus/minus

one standard deviation

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 967–982 (2007)DOI: 10.1002/hyp

976 F. F. WHITAKER AND P. L. SMART

Table III. Calcium carbonate budget for the limestones within the meteoric regime

Zone/process CaXS (mg l�1) CaXS (mg l�1) Flux(m year�1)

MCaCO3

(g m�2 year�1)a (%)b

Rainfall 1·1 ± 0·05 N.A. 1·34 NA NABedrock surface

Undifferentiated bedrock 7·7 ± 2·5 6Ð6 š 2Ð1 1Ð14 19 ± 6·0 NABanana holes 51 ± 19 50 š 20 1Ð14 142 ± 57 NATotal surface dissolutionc 16 š 7Ð9 15 š 7Ð7 1Ð14 43 š 8Ð8 24 š 4Ð9

Vadose zoneEquilibration with CO2 and aragonitec,d 12–24 1Ð14 34–68 19–37

Total surface C vadose zone 28–41 27–40 1Ð14 77–111 42–61Freshwater lens

Mixinge 19–42 �0Ð6 to C1Ð3 1Ð14 �1Ð7 to C3Ð7 0–2Ð0Equilibration with CO2 and aragonitec 54 13–26 1Ð14 37–74 20–41Evaporatione 54 �37 š 6Ð1 1Ð14 �105 š 17 NAOxidation of organic carbon 92 š 12 38 š 12 0Ð34 32 š 10 18 š 5Ð6

Total lens dissolutionf 92 ± 12 51–65 69–110 38–60Net meteoric dissolutionf 92 ± 12 91 ± 12 0·34 77 ± 15 NATotal meteoric dissolutionf NA 128 š 27 NA 182 š 46 100Total meteoric precipitation NA 37 š 6Ð1 NA 105 š 17 NA

a MCaCO3 D CaXS (mg l�1, from Table IV) ð2Ð497 ð flux of water (m year�1, assuming mean annual rainfall (MAR) is 1Ð34 m, interception is15% of MAR and effective recharge is 25% of MAR).b MCaCO3 �%� D MCaCO3 �g m�2 year�1�/

∑meteoric dissolution MCaCO3 �g m�2 year�1� ð 100.

c Range based on 80% rapid runoff, with up to 20% geochemically evolved banana-hole waters.d Range based on 45–75% rapid percolation (see Table II).e Negative numbers indicate loss of calcium from solution due to precipitation.f The large range reflects the possible combinations of rates from different zones/processes, maintaining for equilibration with ground-air CO2 theassociation between high vadose estimates and low phreatic estimates and vice versa.

equilibration with ground-air PCO2 and subsurface car-bonate predominant), the actual figure probably lyingbetween these two extremes. Thus, overall, dissolutionfrom the surface and vadose zone gives a total of27–40 mg l�1 calcium.

Geochemical simulations clearly indicate that mixingof vadose and freshwater lens waters has a minimaleffect on potential dissolution of carbonate. Best esti-mates range from negligible reduction in dissolution atthe water table to minor dissolution within the lens;but in all cases, estimates are within the uncertainty ofmost other estimates of CaXS. However, equilibrationof recharge waters in the upper part of the freshwaterlens with the elevated PCO2 of ground air and aragonitegenerates an additional 13–26 mg l�1 calcium (depen-dent upon proportion of rapid : diffuse percolation). Some60 š 5% of evapotranspiration is derived from the sub-surface in the northern Bahamas; and because of thelimited storage in the vadose zone and shallow depthof the water table in many areas, this is derived mainlyfrom the freshwater lens (Whitaker, 1992). Evaporativeconcentration of water in the upper part of the freshwaterlens while maintaining equilibrium with respect to arag-onite and the PCO2 of ground air results in precipitationof 37 š 6Ð1 mg l�1 calcium, although calcium concentra-tions are maintained at 54 mg l�1 CaXS.

The final process that we consider important in driv-ing carbonate dissolution in the northern Bahamas isoxidation of dissolved organic matter within the freshwa-ter lens. This process likely accounts for the difference

between the observed CaXS in water samples from pump-ing boreholes (92 š 12 mg l�1 calcium) and that mod-elled by a summation of all other processes (54 mg l�1

calcium), i.e. an additional 38 š 12 mg l�1 calcium. Thenet increase in total CaXS in the freshwater lens estimatedfrom modelling is 51–65 mg l�1, which is approximatelydouble the combined surface and vadose estimates. How-ever, accounting for calcium removal due to precipitationdriven by evapotranspiration, the gross increase in CaXS

is 88–102 mg l�1, which is considerably higher than thecombined surface and vadose total.

We now consider the distribution of carbonate dissolu-tion and precipitation within the meteoric zone, this beingthe product of the change in calcium concentration andthe hydrological flux (Table III). Flux totals are based onhydrological data for North Andros (Table I), which hasa climate intermediate between that of the wetter GrandBahama to the north and the drier South Andros. Forsurface, vadose and shallow freshwater lens processesthe hydrological flux is equal to the mean annual rain-fall less interception (average 15% of annual rainfall),but within the freshwater lens it is further reduced byevapotranspiration from the subsurface (average 60% ofannual rainfall). The net flux of carbonate from the mete-oric system is the most well-constrained rate that can beestimated, as it is based solely on the observed calciumexcess in the freshwater lens (92 š 12 mg l�1 calcium)and the annual groundwater flux (0Ð34 m year�1). Thisamounts to CaCO3 of 77 š 15 g m�2 year�1 (Table III).

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 967–982 (2007)DOI: 10.1002/hyp

GEOCHEMISTRY OF METEORIC DIAGENESIS IN CARBONATE ISLANDS: 2 977

However, evapotranspiration results in local reprecipi-tation of CaCO3 of 105 š 17 g m�2 year�1, which isalmost 1Ð4 times greater than the net flux from the mete-oric system. Thus, total meteoric dissolution of CaCO3

is 182 š 46 g m�2 year�1. Of this total, approximatelyequal amounts (between 40 and 60%) are derived fromdissolution above and at/below the water table. Dissolu-tion at the bedrock surface represents just less than one-quarter of the total dissolution, with most of this derivedfrom subsoil dissolution. Note that we have assumed inthese calculations that all non-interception evaporationoccurs from the subsurface. If a proportion of this ismet from surface detention and soil moisture, then thevadose and upper freshwater lens hydrological fluxes willbe reduced accordingly, reducing the total subsurface dis-solution rate. However, the net loss of carbonate is notaffected by this assumption. Similarly, the amount of car-bonate reprecipitation will be reduced, both because ofthe reduced hydrological flux in the upper part of thefreshwater lens and because runoff (but not soil) watershave lower calcium concentrations than waters in thefreshwater lens.

Finally, we can compare the significance of differentsources of CO2 for the dissolution process. Surface runoffand rapid recharge derive CO2 from the atmosphere andare responsible for somewhat more than 8% of totaldissolution (increasing to a maximum of 16% if rapidrecharge equilibrates with aragonite in the subsurface).Dissolution in soil-filled pockets at the bedrock surfaceis driven by soil CO2 and is responsible for up to 16% oftotal meteoric dissolution (CaCO3 of 28 g m�2 year�1).Soil CO2 may also contribute to ground-air CO2, but weaccount for this source separately. Depending on waterrouting in the vadose zone, ground air may be responsiblefor 19–37% of total meteoric dissolution (CaCO3 of34–68 g m�2 year�1). Finally, assuming that organicmatter oxidation occurs after evaporative concentration,18 š 5Ð6% (CaCO3 of 32 š 10 g m�2 year�1) of overalldissolution could result from this process.

DISCUSSION

Evolution of the bedrock surface

Large areas of the subaerial exposure surface in thenorthern Bahamas are characterized by differential relief

generated by dissolution, and seen most markedly inthe distinctive pitted ‘banana hole’ topography. We havepreviously suggested that the pitted topography resultsfrom positive feedbacks that enhance soil PCO2 -drivendissolution within topographic lows compared with theintervening highs (Smart and Whitaker, 1989). Thesefeedbacks include accumulation of organic matter intopographic lows, preferential colonization of vegetationin these accumulating sites and enhanced organic matterdecomposition in deeper, moister soils, all of which leadto higher soil CO2 production.

Rates of surface lowering (SL/mm ka�1) can be calcu-lated from the rate of CaCO3 dissolved (MCaCO3/g m�2

year�1; see Table III), the density of the dissolved car-bonate � (assumed to be calcite, 2Ð71 ð 10�6 g m�3) andfractional porosity of the bedrock n (assumed to be 0Ð3;Halley and Evans, 1983):

SL D MCaCO3

��1 � n��2�

Rapid runoff water has dissolved CaCO3 of 19 š6Ð0 g m�2 year�1, equivalent to a lowering rate for theundifferentiated surface of 10 š 3Ð2 mm ka�1 (Table IV).This may be a slight overestimate due to dissolu-tion of terrestrial dust derived from dry deposition andthe presence of minor amounts of aragonite (�2Ð93 ð10�6 g m�3). A more significant source of uncertainty isthe porosity of the surficial rock. Our assumed fractionalporosity of 0Ð3 is somewhat lower than the 0Ð43 averagequoted by some for Pleistocene limestones (Halley andSchmoker, 1983). A higher porosity would increase sur-face lowering rates by some 20%, but we use a lowervalue in recognition of the enhanced cementation in theshallow subsurface in the northern Bahamas describedby Beach (1995). However, it is comparable to estimatesof surface lowering from bare limestones in northernJamaica (9 mm ka�1; Smith et al., 1972) and from semi-arid regions of South Australia (6–13 mm ka�1; Smithet al., 1995) and rather higher than the figure reported forIsrael (2 mm ka�1, Gierson, 1974).

The surface lowering rate of 75 š 30 mm ka�1 calcu-lated for banana holes (Table IV) is a minimum estimatebecause it ignores concentration of flow into topographiclows on the bedrock surface. Such focusing of over-land flow over short distances is the norm, and mappingdemonstrates that banana holes may have a catchment

Table IV. Rates of surface lowering and porosity modification by meteoric waters in the northern Bahamas. Figures in parenthesesrefer to evaporation of diffuse percolation in the vadose zone

Zone Surface lowering (mm ka�1) Porosity generation(% ka�1)

Porosity loss(% ka�1)

Average bedrock surface 17–24 NA NAUndifferentiated bedrock 10 š 3Ð2 NA NABanana holes 75–150 NA NA

Average vadose zone NA 1Ð6–3Ð2 (0Ð3–1Ð0) ND (1Ð2–2Ð7)Average freshwater lens NA 0Ð25–0Ð40 (0Ð29–0Ð48) 0Ð38 š 0Ð06 (0Ð2–0Ð4)

Lens top (z < 1 m) NA 1Ð4–4Ð0 (1Ð8–4Ð9) 4Ð0 š 0Ð64 (1Ð8–3Ð0)Bulk of lens (z > 1 m) NA 0Ð064 NA

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 967–982 (2007)DOI: 10.1002/hyp

978 F. F. WHITAKER AND P. L. SMART

area at least twice their surface area (Smart and Whitaker,unpublished data). This could increase the surface low-ering rate in banana holes to at least 150 š 60 mm ka�1,some 15 times that of the surrounding undifferentiatedbedrock surface. Using the minimum rate, the deepestbanana holes observed on Andros (¾5 m deep) coulddevelop from undifferentiated bedrock in 67 š 27 ka,reducing to 33 š 13 ka where focusing of overland flowis assumed. These estimates are well within the estimatedage of the youngest rocks (125 ka; Carew et al., 1998)in which the banana holes are developed.

Marine erosion during periods of rising sea level islikely to alter these subaerial exposure surfaces consider-ably (Rasmussen and Neumann, 1988), but rapid marine(or prior subaerial) sedimentation could preserve the dif-ferentiated topography. Pits of similar dimensions andspacing to the Bahamian banana holes have been reportedfrom the Carboniferous Limestone of the UK (Walkden,1974; Adams and Horbury, 1989; Vanstone 1998). Usingour differential erosion estimate, the depth of these fea-tures could provide an order-of-magnitude estimate of theduration of exposure, even if the palaeoclimate is not wellconstrained.

Porosity development in the vadose zone and freshwaterlens

Depth-averaged rates of porosity generation n (%ka�1) can be calculated from rates of CaCO3 dissolutionMCaCO3 (g m�2 year�1; see Table IV), mineral density� (g m�3) and the average thickness z (m) of thehydrological zone:

n D MCaCO3

�zð 100 �3�

For an average vadose zone depth of 0Ð8 m (NorthAndros), CaCO3 dissolution of 34–68 g m�2 year�1

(Table IV) generates porosity at a rate of 1Ð6–3Ð2%ka�1 (Table IV). Over the 125 ka since deposition of theyoungest rocks, this would cause a porosity increase of2Ð0–4Ð0%.

The rate of total dissolution of CaCO3 in the freshwaterlens is estimated to be 69–110 g m�2 year�1 (Table IV).Assuming an average freshwater lens thickness of 10Ð5 m(North Andros, Table I), this is equivalent to an aver-age rate of porosity generation of 0Ð25–0Ð40% ka�1

(Table IV). Thus, the average rate of porosity generationin the vadose zone is 4–13 times greater than the aver-age for the freshwater lens. This significant differenceis related primarily to the order-of-magnitude differencein average thickness of the zones. However, net ratesof porosity creation in the freshwater lens are reducedby reprecipitation of some of the carbonate dissolved(CaCO3 of 105 š 17 g m�2 year�1). Were this precipi-tation to be uniformly distributed through the freshwaterlens, the rate of porosity occlusion by cementation wouldbe 0Ð38 š 0Ð06% ka�1, approximately equal to or some-what greater than the rate of average dissolution.

However, neither dissolution nor precipitation is likelyto be uniformly distributed within the lens. Reprecipi-tation of carbonate is likely concentrated in the zonesubject to evaporative concentration at the water table.The depth of this zone is poorly constrained, but here weassume that all reprecipitation occurs within the uppermetre of the freshwater lens (approximately the depth ofthe zone occupied by the annual effective recharge). Thisassumption is supported by observations of the distribu-tion of cements (Budd, 1984) and measurements of min-eral transformation (McClain et al., 1992) in Holoceneaquifers. The rate of porosity occlusion driven by evapo-ration from this upper freshwater lens zone would, there-fore, average some 4Ð0 š 0Ð64% ka�1. In this zone, therate of cementation is at least 10 times that of dissolution(assuming the dissolution rate is homogeneous through-out the freshwater lens). This rate would be rather lowerif some of the evaporation occurs from the vadose zone.

Dissolution is also likely to be greater in this upperpart of the freshwater lens. In particular, dissolution byundersaturated vadose recharge (which has equilibratedwith ground-air CO2) would occur once the watersreached the top of the freshwater lens and the residencetime exceeded the time for full equilibration of theaqueous and solid carbonate phases. This would generate1Ð4–2Ð8% ka�1 porosity (assuming the same 1 m thickzone below the water table). Unless organic matteroxidation is also focused almost entirely within theshallowest part of the freshwater lens (which wouldgenerate a maximum total rate of porosity generation of4Ð0% ka�1 assuming the same 1 m thick active zone),the overall implication is that there is likely to be azone of net cementation at the water table. Althoughcementation by calcite may dominate volumetrically, thiswill be accompanied by coeval active dissolution ofmetastable allochems. Inversion will occur from primaryinterparticle porosity to intra-particle secondary mouldic,vug and channel porosity. With rapid transmission ofmore aggressive waters and surface-derived organics viathe more transmissive components of the pore network,positive feedbacks will continue to enlarge these at theexpense of the surrounding rock.

At greater depth within the freshwater lens, porositygeneration may continue driven by organically mediatedreactions; but, with dissolutional potential distributedover a thicker zone, this will occur at a very much slowerrate (up to 0Ð12% ka�1). This is some 13–26 times lowerthan the rate of porosity generation in the vadose zone. Infact, dissolution probably decreases systematically withdepth, with renewal of dissolution potential at redoxinterfaces as well as in the underlying fresh-water–salt-water mixing zone (Smart et al., 1988; Whitaker andSmart, 1997c).

In the more arid northwestern Yucatan, Mexico, activeprecipitation of cements at the water table has beenreported by Perry et al. (1989). Their volume is suffi-cient to occlude all porosity, creating a low-permeabilityaquiclude at the water table that confines the underly-ing freshwater lens. In the northern Bahamas, drillers

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 967–982 (2007)DOI: 10.1002/hyp

GEOCHEMISTRY OF METEORIC DIAGENESIS IN CARBONATE ISLANDS: 2 979

report a particularly well lithified zone at depth withinthe modern freshwater lens, which can create semi-confined conditions (Little et al., 1977). In cores fromNorth Andros, Beach (1995) also reports the pres-ence of cemented horizons where primary porosity isoccluded by phreatic freshwater cements, but signifi-cant secondary (dissolutional) porosity remains. Beach(1995) associates these zones with unconformities, andwe suggest that this could arise from elimination ofthe vadose zone by surface lowering prior to accu-mulation of sediment during the following sea-levelhigh-stand. Such a suggestion is supported by ourrate data, some 80 ka of glacial low-stand loweringbeing required to totally eliminate the rock withinthe present day (0.8 m deep) vadose zone, based onlowering of the undifferentiated bedrock surface at10 mm ka�1.

In our modelling we have assumed that evapotran-spirative demand is satisfied wholly by interception anddirect loss from the water table. In some areas that havedeeper vadose zones, e.g. in the aeolianite ridges alongthe east coast of Andros, some evapotranspiration maybe sourced from water retained in the vadose zone. Ifwe assume that the 25–55% of rainfall that fails toreach the water table within 12 h of a storm evapo-rates from the vadose zone (Table IV), then the result-ing cementation could occlude porosity in the vadosezone at a rate of 1Ð2–2Ð7% ka�1. There will also bea reduction of dissolutional porosity generation in thevadose zone to 0Ð3–1Ð0% ka�1, because of the lowerhydrological flux. This would then result in net porosityocclusion in the vadose zone, and provides an alterna-tive possible explanation for the reduction in cementationwith depth below the unconformity observed by Beach(1995). Should this occur, then evaporation will prefer-entially cement the finer pores that retain and transmitwater by capillarity (as observed by Brooks and Whitaker(1997) for Holocene sands on Grand Bahama). How-ever, this model seems unlikely, as Beach (1995) reportsa predominance of phreatic (equant spar and micro-spar) cements and relative scarcity of vadose (menis-cus, micrite and needle-fibre) cements. Therefore, webelieve that our original model, although poorly con-strained by hydrological observations, matches rather bet-ter the longer term diagenetic products observed in thefield area.

The total amount of CaCO3 dissolution from the mete-oric systems of North Andros can also be calculatedfrom the product of the area of the island occupiedby each hydrological zone (Table I) and the rate ofdissolution per square metre in that zone (Table III).Over the whole of North Andros (area 3260 km2), atotal of �2Ð5–3Ð6� ð 1011 g of CaCO3 is dissolved annu-ally from the surface and vadose zone. Only 42% ofthe land area is underlain by the freshwater lens (Cantand Weech, 1986). Over the remaining area, the mix-ing zone extends to the water table. Total dissolutionof CaCO3 within the freshwater lens thus amounts to�1Ð1–1Ð5� ð 1011 g year�1. This is less than half the

total mass of carbonate dissolved from surface andvadose zones. The total export of CaCO3 each year tothe oceans from this one island due to meteoric dis-solution thus amounts to �3Ð6–5Ð1� ð 1011 g. Dissolu-tion within the meteoric system on North Andros willalso sequester atmospheric CO2 carbon, equivalent to�4Ð3–6Ð1� ð 1010 g year�1 of carbon. This is likely anunderestimate of total sequestration, as it excludes thesignificant additional dissolution in the fresh-water–salt-water mixing zone.

Comparison with dissolution rates on other carbonateislands

There have been relatively few hydrochemical studiesof carbonate islands that allow estimation of present-dayrates of surface lowering and/or subsurface porosity gen-eration (Table V, expanded and modified from Anthonyet al. (1989)). The average concentration of dissolvedcarbonate rock (calculated from CaXS and also MgXSfor islands where high-magnesium calcite is present) inthe freshwater lens reported in these studies ranges from2Ð5 ð 10�4 kg l�1 on Water Cay to 4Ð0 ð 10�4 kg l�1 onEnewetak. These concentrations are independent of islandarea, but correlate strongly with recharge, which is alarger fraction of rainfall in the wetter, more equatorialPacific islands than in the drier semi-arid islands, such asthose in the Bahamas (Figure 7). The exception to this isOcean Bight, where carbonate dissolution is almost twicethat predicted from the general relationship, possiblyreflecting the unusually abundance of high-magnesiumcalcite, which is more reactive. The climatic control ondissolution rates is strongly supported by the variation inhydraulic conductivity (secondary permeability) as mea-sured by pumping tests across the climatic gradient inthe Bahamian Archipelago (Whitaker and Smart, 1997a).Hydraulic conductivity averages over two orders of mag-nitude higher in the wetter northern Bahamas (meanannual rainfall: 1550 mm) than in the more arid southernislands (mean annual rainfall: 810 mm).

On Holocene islands dominated by metastable miner-alogies, this infers less efficient stabilization of arago-nite (sensu Budd, 1988) with less local reprecipitationof dissolved aragonite as low-magnesium calcite cement,and possibly also higher overall rates of stabilization onislands with a wetter climate. This suggestion is sup-ported by geological studies that have quantified variationin the abundance of carbonate minerals in deposits of dif-ferent age. For example, on Barbados (Harrison, 1975),high-magnesium calcite is still present after 300 ka onthe drier south coast, whereas stabilization is completewithin 83 ka on the wetter west coast.

Higher recharge gives greater fluid fluxes; thus, disso-lution rates are much higher on wet islands (355 m3 km�2

year�1 on Laura, precipitation 1780 mm year�1) than dryislands (38 m3 km�2 year�1 on Water Cay, precipitation285 mm year�1). However, this climatic control may alsooperate via the effect of rainfall and temperature on veg-etation and soil development. Thus, wetter islands withmore vegetation and thicker soils would have a higher

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 967–982 (2007)DOI: 10.1002/hyp

980 F. F. WHITAKER AND P. L. SMART

Table V. Comparison of calculated dissolution rates for carbonate islands

Island NorthAndros

Bermudaa Laurab Enewetakb WaterCayc

OceanBightd

Low-Mg calcite (%) 50–100 40–60 40–70 30–100 <5 <1Latitude (°N) 25 32 7 12 25 24Lens area (km2) 1370 14 1Ð4 1Ð06 0Ð025 1Ð5Rainfall (mm year�1) 1340 1460 3560 1470 1141e 1000Recharge (mm year�1) 339 370 1780 500 285f 200Flushing (l year�1) 4Ð64 ð 1011 5Ð2 ð 109 2Ð5 ð 109 5Ð0 �3Ð0� ð 108h 7Ð13 ð 104 3Ð0 ð 108

CaCO3 �10�4 kg l�1� 2Ð3 3Ð3 3Ð7 4Ð0 2Ð5 3Ð8Dissolution (kg year�1) 1Ð1 ð 108 1Ð7 ð 106g 9Ð2 ð 105 2Ð1 �1Ð2� ð 105h 1Ð8 ð 101 1Ð13 ð 105

Volume dissolved(m3 year�1)i 5Ð6 ð 105 9Ð0 ð 102 4Ð9 ð 102 1Ð1 �0Ð6� ð 102h 9Ð4 ð 10�3 5Ð9 ð 101

(m3 km�2 year�1) 41 64 347 105 (61)h 38 40

a Data from Plummer et al. (1976).b Data from Anthony et al. (1989).c Data from Budd (1988).d Data from McClain et al. (1992).e Based on total for Eleuthera (Bahamas Met Office).f Based on effective recharge of 25% of total rainfall.g Ten times higher than in Anthony et al. (1989) probably due to typographic error in original table.h Product of area and recharge indicates flushing of 0Ð5 ð 109 l year�1, not 0Ð3 ð 109 in original table (figures based on original calculation offlushing included in brackets).i Assuming density of 2Ð71 g cm�3 and porosity of 30%.

Ocean Bight

Bermuda

Enewetak

N Andros

Water Cay

Laura1,780mm/yr−1

2002

3

4

400 600Effective Recharge mm.yr −1

CaC

O3

(kg.

L−1 ×

10−4

)

Figure 7. Relationship between effective recharge and calcium carbonatedissolution for carbonate islands (see Table V). Note recharge for Laura

is off scale

soil and ground-air CO2 and, consequently, higher cal-cium concentrations. CaCO3 concentrations on Laura areslightly lower than on Enewetak, despite the much higherrecharge on Laura. This suggests that, rather than a directrelationship with fluid flux, soil PCO2 may be the criticalcontrol, and that, above ¾500 mm year�1, recharge nolonger limits soil PCO2 .

SUMMARY AND IMPLICATIONS FORCARBONATE DIAGENESIS

Geochemical modelling in combination with field watersampling has provided a detailed and quantitative anal-ysis of the magnitude of dissolution and precipitationprocesses in the freshwater systems of the Pleistocenecarbonate islands of the northern Bahamas. The resultsreveal dissolution extending from the land surface to deepwithin the freshwater lens, but with the greatest mass

transfers occurring at the top of the lens, in the vadosezone, and in soil-filled banana holes on the bedrock sur-face.

Runoff from the predominately bare or litter-covered(undifferentiated) bedrock surface has little contact withsoil and, thus, equilibrates with atmospheric PCO2 , butresidence times are too short for equilibration with arago-nite. Rates of surface dissolution, therefore, are relativelylow, as has been reported from other arid and semi-aridcarbonate terrains (e.g. Smith et al., 1995). Some sur-face waters are channelled into topographically definedbedrock hollows (banana holes) that occupy ¾10% ofthe surface, and within these contact organic soils withelevated PCO2 . Dissolution rates for banana holes are7Ð5–15 times those for undifferentiated bedrock, gener-ating small-scale relief at a rate of 65–140 mm ka�1 anda distinctive pocketed topography formed by differentialdissolution. Such relief, which cannot be adequately sam-pled at core scale (Budd et al., 2002), has potential as anindicator of the duration of exposure or palaeoclimate ofexposure surfaces in the geological record.

A body of ground air with elevated PCO2 appears tocontrol the evolution of both diffuse percolation andwater at the top of the freshwater lens. CO2 is supplied bydownward diffusion from the soil and degassing from thewater table. Within the vadose zone, the lowest predictedrate of dissolution (porosity of 1Ð6% ka�1) occurs withrapid percolation of more than half the recharge waters(scenario B.1), but all other percolation scenarios give adissolution rate approximately double this rate (porosityof 27–32% ka�1). Total vadose dissolution is approx-imately equivalent to the amount of dissolution in theupper lens from equilibration of percolation waters withground air and aragonite. Within the freshwater lens, CO2

generated by oxidation of organic matter, predominately

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 967–982 (2007)DOI: 10.1002/hyp

GEOCHEMISTRY OF METEORIC DIAGENESIS IN CARBONATE ISLANDS: 2 981

(but not exclusively) by aerobic processes, is responsiblefor driving almost 20% of total dissolution, the remainderbeing driven by surface-derived carbonic acid. In termsof carbonate fluxes, mixing processes are insignificantwithin the meteoric system of the northern Bahamas.

Because the vadose zone on Andros is very thin,direct evapotranspiration from the top of the lens drivesredeposition of some of the dissolved carbonate. Wherediffuse vadose percolation dominates, redeposition nearthe water table could result in net porosity occlusion at arate of up to 2Ð6% ka�1. However, equilibration of rapidvadose recharge and oxidation of surface-derived organicmatter together may generate porosity of 1Ð4–4Ð0% ka�1,which is approximately equal to rates of precipitationdriven by evaporation (4Ð0 š 0Ð64% ka�1). Thus, thereis a very significant dissolution of metastable primarycarbonate and cementation with calcite, with inversionof primary intergranular to secondary mouldic, vugand karstic porosity, but little change in total porosity.The percentage of total dissolution occurring within thefreshwater lens is comparable to that from the surfaceand vadose zone combined (Table III). However, averagedissolution rates within the freshwater lens are an order ofmagnitude lower than those in the vadose zone, directlyreflecting differences in zone thickness. The upper partof the freshwater lens is thus distinct, not only as a focusfor dissolution (with porosity generation at rates 12–33times those of the deeper lens), but also for precipitationof cements by evaporative concentration.

Two factors control the distribution and rates of mete-oric diagenesis: hydrological routing and depth of thevadose zone. Our modelling of current vadose processesshows that the proportion of rapid to diffuse percola-tion is critical in controlling distribution of dissolution.On initial exposure, diffuse intergranular flow will dom-inate and the vadose dissolution rate will exceed that inthe upper freshwater lens. However, capillary retention ofpercolation will also be greater, enhancing vadose cemen-tation, particularly in fine-grained sediments. With time,vadose dissolution will reduce as recharge is increas-ingly routed down dissolutionally enlarged preferentialflow paths, enhancing dissolution just below the watertable. Vadose zone depth is controlled by relative sealevel. For Andros, a sea level fall of only a few metreswould decouple the saturated zone from evapotranspira-tive loss. Assuming no change in the hydrology of thevadose zone, precipitation driven by evaporation wouldthen be focused wholly in the vadose zone and be signif-icantly reduced, giving an increased effective recharge.The resulting higher groundwater flux in the saturatedzone may increase the dissolution rate in the diagenet-ically active freshwater lens top zone and cementationdriven by evaporation from the water table would cease.

With a larger drop in sea level (typical of Pleis-tocene low-stands), dissolution potential would becomedistributed over a greater depth of vadose zone and therate of porosity increase would be reduced. More impor-tant, water reaching the water table will be more likelyto have reached equilibrium with ground-air PCO2 and

aragonite/calcite. Thus, dissolution in the lens will begreatly reduced. This effect will be compounded by reten-tion and oxidation of percolating organic matter withinthe very much thicker vadose zone. Precisely this effectis hypothesized by Melim (1996) to explain the appar-ent absence of diagenesis within the low-stand lens onthe western Great Bahama Bank. However, this effect isstrongly dependent upon the diagenetic maturity of thecarbonates. For example, in the Miocene to Pleistocenecarbonates of Guam, storm water rapidly recharges thewater table through a 60–180 m thickness vadose zone(Jocson et al., 2002), resulting in significant dissolutionwithin the freshwater lens (Whitaker et al., 2006). Thissuggests that, contrary to Melim’s (1996) model of a dia-genetically inert low-stand freshwater lens, most icehouseplatforms would experience active phreatic meteoric dia-genesis during major sea level low-stands, as well asduring exposure by minor sea level falls occurring withinhigh-stand periods.

In conclusion, we believe that a clearer understandingof the hydrology of recharge and evapotranspirationprocesses (particularly in the vadose zone) and theresulting geochemical evolution of vadose recharge wateris needed to understand better the controls on rate anddistribution of diagenesis within carbonate islands. Theorigin of meteoric cements following completion ofstabilization is poorly understood from the geologicalrecord (Budd et al., 1993). Here, for the first time, wesuggest that precipitation of phreatic meteoric cementscan be driven by evaporative losses associated with a thinvadose zone and is focused in a limited active zone nearthe water table. Where the vadose zone is thicker, this willcease and the freshwater lens will become predominantlya zone of dissolution even in a semi-arid climate.

ACKNOWLEDGEMENTS

Fieldwork was supported logistically by Steve Hobbsand Neil Sealey, and financially by Exxon Produc-tion Research, Amoco Production Company and ShellKSEPL. We thank Drew Ellis for cartography and theGovernment of the Commonwealth of the Bahamas forpermission to undertake this research. This paper bene-fited from thorough and thought-provoking review by theanonymous reviewers.

REFERENCES

Adams AE, Horbury AD. 1989. Tree root structures in a Dinantianpaleokarst, Urswick Limestone, south Cumbria. Proceedings of theYorkshire Geological Society 47: 345–348.

Anthony SS, Peterson FL, Mackenzie FT, Hamlin SN. 1989. Geohydrol-ogy of the Laura fresh-water lens, Majuro atoll: a hydrogeochemicalapproach. Geological Society of America Bulletin 101: 1066–1075.

Balek J. 1983. Hydrology and Water Resources in Tropical Regions .Elsevier: Amsterdam.

Beach DK. 1995. Controls and effects of subaerial exposure oncementation and development of secondary porosity in the sub-surface of the Great Bahama Bank. American Association of PetroleumGeologists Memoir 63: 1–33.

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 967–982 (2007)DOI: 10.1002/hyp

982 F. F. WHITAKER AND P. L. SMART

Beach DK, Ginsburg RN. 1980. Facies succession of Pliocene–Pleistocene carbonates, northwestern Great Bahama Bank. AmericanAssociation of Petroleum Geologists Bulletin 64: 1634–1641.

Bogli A. 1964. Corrosion par melange des eaux. International Journal ofSpeleology 1: 61–70.

Brooks SM, Whitaker FF. 1997. Co-evolution of early diagenesis andvadose zone hydrology of Holocene carbonate sands. Earth SurfaceProcesses and Landforms 22: 45–58.

Budd DA. 1984. Freshwater diagenesis of Holocene ooid sands,Schooner Cays, Bahamas . PhD thesis, University of Texas at Austin(unpublished).

Budd DA. 1988. Aragonite-to-calcite transformations during fresh-water diagenesis of carbonates: insights from pore-water chemistry.Geological Society of America Bulletin 100: 1260–1270.

Budd DA. 2001. Permeability loss with depth in the Cenozoic carbonateplatform of west-central Florida. American Association of PetroleumGeologists Bulletin 85: 1253–1272.

Budd DA, Hammes U, Vacher HL. 1993. Calcite cementation in theupper Floridan aquifer: a modern example for confined-aquifercementation models? Geology 21: 33–36.

Budd DA, Gaswith SB, Oliver WL. 2002. Quantification of macroscopicsubaerial exposure features in carbonate rocks. Journal of SedimentaryResearch 72: 917–928.

Campbell DG. 1978. The Ephemeral Isles . MacMillan: London.Cant RV, Weech PS. 1986. A review of the factors affecting the

development of Ghyben–Herzberg lenses in the Bahamas. Journal ofHydrology 84: 333–343.

Carew JL, Mylroie JE, Schwabe S. 1998. The geology of South Andros:a reconnaissance report. Cave and Karst Science 25: 57–66.

Crockford RH, Richardson D. 1990. Partitioning of rainfall in aEucalyptus forest and pine plantation in southeastern Australia.Hydrological Processes 4: 131–188.

Falkland AC, Brunel JP. 1993. Review of the hydrology and waterresources of humid tropical islands. In Hydrology and WaterManagement in the Humid Tropics , Bonell M, Hufschmidt MM,Gladwell JS (eds). Cambridge Press: Cambridge; 135–163.

Garrett P, Gould SJ. 1984. Geology of New Providence Island, Bahamas.Geological Society of America Bulletin 95: 209–220.

Gierson R. 1974. Karst processes in the eastern Upper Galilea, northernIsrael. Journal of Hydrology 21: 131–152.

Halley RB, Evans CC. 1983. The Miami Limestone: a guide to selectedoutcrops and their interpretation. Miami Geological Society: Miami.

Halley RB, Schmoker JW. 1983. High-porosity Cenozoic carbonate rocksof south Florida: progressive loss of porosity with depth. AmericanAssociation of Petroleum Geologists Bulletin 67: 191–200.

Harris JG, Mylroie JE, Carew JL. 1995. Banana holes—unique karstfeatures of the Bahamas. Carbonates and Evaporites 10: 215–224.

Harrison RS. 1975. Near-surface subaerial diagenesis of Pleistocenecarbonates, Barbados, West Indies . PhD thesis, Brown University(unpublished).

Jocson JMU, Jenson JW, Contractor DN. 2002. Recharge and aquiferresponse: Northern Guam Lens Aquifer, Guam, Mariana Islands.Journal of Hydrology 250: 231–254.

Land LS, Mackenzie FT, Gould SJ. 1967. Pleistocene history ofBermuda. Geological Society of America Bulletin 78: 993–1006.

Little BG, Buckley DK, Jefferiss A, Stark J, Young RN. 1973. LandResources of the Commonwealth of the Bahamas; Volume 3, AndrosIsland . Ministry for Overseas Development: Surbiton, UK.

Little BG, Buckley DK, Jefferiss A, Stark J, Young RN. 1975. LandResources of the Commonwealth of the Bahamas, Volume 5, GrandBahama. Ministry for Overseas Development: Surbiton, UK.

Little BG, Buckley DK, Cant RV, Henry PWT, Jefferiss A, Mather JD,Stark J, Young RN. 1977. Land Resources of the Bahamas: Summary .Ministry for Overseas Development: Surbiton, UK.

McClain ME, Swart PK, Vacher HL. 1992. The hydrochemistry of earlymeteoric diagenesis in a Holocene deposit of biogenic carbonates.Journal of Sedimentary Petrology 62: 1008–1022.

Melim LA. 1996. Limitations on lowstand meteoric diagenesis in thePliocene–Pleistocene of Florida and Great Bahama Bank: implicationsfor eustatic sea-level models. Geology 24: 893–896.

Moore CH. 2001. Carbonate Reservoirs—Porosity Evolution andDevelopment in a Sequence Stratigraphic Framework . ElsevierScience: Amsterdam.

Morse JW, Mackenzie FT. 1990. Geochemistry of Sedimentary Carbon-ates . Developments in Sedimentology 48. Elsevier Science: Amster-dam.

Mylroie JE, Carew JL. 1990. The flank margin model for dissolutionalcave development in carbonate platforms. Earth Surface Processes andLandforms 15: 413–424.

Mylroie JE, Carew JL. 1995. Geology and karst geomorphology of SanSalvador Island, Bahamas. Carbonates and Evaporites 10: 193–206.

Parkhurst DL. 1995. User’s guide to PHREEQC—a computer programfor speciation, reaction-path, advection-transport and inverse geochem-ical calculations . United States Geological Survey Water ResourcesInvestigation Report 95–4227.

Perry EC, Swift J, Gamboa J, Reeve A, Sanborn R, Marın LE, Villa-suso M. 1989. Geologic and environmental aspects of surface cemen-tation, north coast, Yucatan, Mexico. Geology 17: 818–821.

Plummer LN. 1977. Defining reactions and mass transfer in part of theFloridian aquifer. Water Resources Research 13: 801–812.

Plummer LN, Busenberg E. 1982. The solubilities of calcite, aragoniteand vaterite in CO2 –H2O solutions between 0° and 90 °C, and anevaluation of the aqueous model for the system CaCO3 –CO2 –H2O.Geochimica et Cosmochimica Acta 46: 1011–1040.

Plummer LN, Vacher HL, Mackenzie FT, Bricker OP, Land LS. 1976.Hydrogeochemistry of Bermuda: a case history of ground-waterdiagenesis of biocalcarenites. Geological Society of America Bulletin87: 1301–1316.

Rasmussen KA, Neumann AC. 1988. Holocene overprints of Pleistocenepaleokarst, Bight of Abaco, Bahamas. In Paleokarst , James NP,Choquette PW (eds). Springer-Verlag: New York; 132–148.

Rossinski V, Wanless R. 1992. Topographic and vegetative controls oncalcrete formation, Turks and Caicos Islands, British West Indies.Journal of Sedimentary Petrology 62: 84–98.

Runnels DD. 1969. Diagenesis, chemical sediments, and the mixing ofnatural waters. Journal of Sedimentary Petrology 39: 1188–1201.

Sanford W, Konikow L. 1989. Porosity development in coastal carbonateaquifers. Geology 17: 249–252.

Smart PL, Whitaker FF. 1989. Controls on the rate and distribution ofcarbonate bedrock dissolution in the Bahamas. In Proceedings, 4thSymposium on the Geology of the Bahamas, College of the Finger LakesField Study Centre, San Salvador, Bahamas; 213–221.

Smart PL, Dawans JM, Whitaker FF. 1988. Carbonate dissolution in amodern mixing zone. Nature 335: 811–813.

Smith DI, Drew DP, Atkinson TC. 1972. Hypothesis of karst develop-ment in Jamaica. Transactions of the British Cave Research Group 14:159–173.

Smith DI, Greenaway MA, Moses C, Spate AP. 1995. Limestoneweathering in eastern Australia. Part 1: erosion rates. Earth SurfaceProcesses and Landforms 20: 451–463.

Thrailkill J. 1968. Chemical and hydrologic factors in excavation oflimestone caves. Geological Society of America Bulletin 79: 19–46.

Vacher HL, Bengtsson TO. 1989. Effect of hydraulic conductivity onthe residence time of meteoric ground water in Bermudian- andBahamian-type islands. In Proceedings, 4th Symposium on the Geologyof the Bahamas, College of the Finger Lakes Field Study Centre, SanSalvador, Bahamas; 337–351.

Vanstone SD. 1998. Late Dinantian paleokast of England and Wales;implications for exposure surface development. Sedimentology 45:1–19.

Walkden GM. 1974. Paleokarstic surfaces in Upper Visean (Carbonifer-ous) limestones of the Derbyshire Block, England. Journal of Sedimen-tary Petrology 44: 1232–1247.

Whitaker FF. 1992. Hydrology, geochemistry, and diagenesis of moderncarbonate platforms in the Bahamas . PhD thesis, University of Bristol(unpublished).

Whitaker FF, Smart PL. 1997a. Climatic control of hydraulic conductiv-ity of Bahamian limestones. Ground Water 35: 859–868.

Whitaker FF, Smart PL. 1997b. Hydrogeology of the BahamianArchipelago. In Geology and Hydrology of Carbonate Islands ,Vacher HL, Quinn T (eds). Elsevier Science: Amsterdam; 183–216.

Whitaker FF, Smart PL. 1997c. Groundwater circulation and geochem-istry of a karstified bank-marginal fracture system, South AndrosIsland, Bahamas. Journal of Hydrology 197: 293–314.

Whitaker FF, Smart PL. 1998. Hydrology, geochemistry and diagenesisof fracture blue holes, South Andros. Cave and Karst Science 25:75–82.

Whitaker FF, Smart PL. 2007. Geochemistry of meteoric diagenesis incarbonate islands of the northern Bahamas: 1. Evidence from fieldstudies. Hydrological Processes 21: this issue.

Whitaker FF, Smart PL, Hague Y, Waltham DA, Bosence DJW. 1997.A coupled two-dimensional sedimentological and diagenetic model forcarbonate platforms. Geology 25: 175–178.

Whitaker FF, Johnson VJ, Paterson RJ. 2006. Meteoric diagenesis duringsea-level lowstands: evidence from modern hydrochemical studies onnorthern Guam. Journal of Geochemical Exploration 89: 420–423.

Copyright 2007 John Wiley & Sons, Ltd. Hydrol. Process. 21, 967–982 (2007)DOI: 10.1002/hyp