Ground truthingchemostratigraphiccorrelations in fluvial systemsK. T. Ratcliffe, A. Wilson, T. Payenberg,A. Rittersbacher, G. V. Hildred, and S. S. Flint

ABSTRACT

Changes in elemental chemistry have been used to define strati-graphic correlations between wellbores in petroleum basins. Fewpublications, however, relate defined chemical stratigraphy tophysical correlations, and none have been found that do so in flu-vial systems. Here, chemostratigraphy is applied to Permian flu-vial sediments within the Beaufort Group of the Karoo Basin inSouth Africa, and a correlation between three logged sections isdefined. This correlation is tested against physically determinedchronostratigraphic correlations achieved using Heli-LIDAR datato provide a high-resolution correlation between two sections7 km (4.4 mi) apart, and mapping of strata using Google Earth toproduce a correlation between sections 25.5 km (15.8 mi) apart.

The chemostratigraphic characterization that is defined usingdata from fine-grained lithologies resulted in the recognition ofeight chemostratigraphic packages, with thicknesses between 50and 250 m (164 and 820 ft) over a stratigraphic interval of approx-imately 900 m (2953 ft). Two distinctive changes in geochemicalcomposition of the coarser lithologies (fluvial channel belts) wereseen over this interval. In the two sections that are 7 km (4.4 mi)apart, higher resolution subdivision of chemostratigraphic pack-ages was achieved to produce four correlative geochemical units(30–60 m [98–197 ft] in thickness) that provide a high-resolutioncorrelation.

The chemostratigraphic and chronostratigraphic correlationsare in close agreement in both the 7-km- (4.4-mi) and the 25.5-km- (15.8 mi) spaced sections. The thickness of the study intervaland spacing of sections is analogous to published chemostratigra-phy studies on subsurface sequences; thereby, ground truthing theuse of chemostratigraphy for correlation in subsurface fluvial

Copyright ©2015. The American Association of Petroleum Geologists. All rights reserved.

Manuscript received July 11, 2013; provisional acceptance December 4, 2013; revised manuscriptreceived March 31, 2014; final acceptance June 5, 2014.DOI: 10.1306/06051413120

AUTHORS

K. T. Ratcliffe ∼ Chemostrat Ltd., Unit 1,Ravenscroft Court, Buttington Cross EnterprisePark, Welshpool, Powys SY21 8SL, UnitedKingdom; kenratcliffe@chemostrat.com

Ken Ratcliffe, a director of Hafren Scientific, worksfor the Chemostrat group of companies. Prior tocofounding Chemostrat Ltd. in 1994, Ken gaineda B.Sc. degree in geology from Imperial College,London, (1984) and a Ph.D. from AstonUniversity, Birmingham, United Kingdom (1987).He then worked as a lecturer at the University ofKingston-upon-Thames before moving into theservice sector in 1989.

A. Wilson ∼ University of Manchester, School ofEarth, Atmospheric and Environmental Sciences,United Kingdom; Present address: ChemostatAustralia Pty, Unit 12, 33 Delawney Street,Balcatta, WA 6021, Australia; aw259@cantab.net

Andy Wilson is the geological manager forChemostrat Australia. Prior to joining Chemostrat,Andy was a researcher at the University ofLiverpool and the University of Manchester. Andyholds B.A. (with honors) and M.Sci. degrees and aPh.D. in geology from the University ofCambridge, United Kingdom.

T. Payenberg ∼ Chevron Energy TechnologyCompany, 250 St Georges Terrace, Perth,Western 6000, Australia; tobi.payenberg@chevron.com

Tobi Payenberg is a stratigrapher in the ClasticStratigraphy R&D Team with Chevron EnergyTechnology Company, researching stratigraphyand reservoir architecture of fluvial and shallowmarine deposits. Prior to Chevron, Tobi was anassociate professor at University of Adelaide. Heholds an M.Sc. degree from the QueenslandUniversity of Technology, Australia, and a Ph.D.from the University of Toronto, Canada.

A. Rittersbacher ∼ UniCIPR, Centre forIntegrated Petroleum Research, Allégaten 41,N-5077 Bergen, Norway; ARITT@statoil.com

Andreas Rittersbacher obtained his diplomadegree in geology/sedimentology from theUniversity of Heidelberg, Germany (2009), andhis Ph.D. from the University of Bergen, Norway(2013) for a study on sedimentary architecture influvial sandstones in Utah and South Africa. He iscurrently working as senior geologist for Statoil intheir technology excellence group.

AAPG Bulletin, v. 99, no. 1 (January 2015), pp. 155–180 155

systems that are, to some degree, analogous to the Beaufort Groupsediments of this paper.

INTRODUCTION

The technique of elemental chemostratigraphy (the use of changesin elemental composition determined from whole rock inorganicgeochemistry to define stratigraphic frameworks) has been usedfor several decades. However, its use in the oil and gas industryhas proliferated over the past decade (Pearce et al., 2005a, b;Ratcliffe et al., 2006; Hildred et al., 2010; Ratcliffe et al., 2010;Wright et al., 2010; Ratcliffe et al., 2012a, b) as a result of thegrowing realization of the independent and objective characteriza-tion and correlation it can provide. Because of its objectivity andindependence, it is commonly employed to provide an alternativesolution where other stratigraphic methods are providing conflict-ing models. Typically, more recent publications use elementalchemostratigraphy on data derived from wellbore core and cut-tings samples. The resultant elemental data have been used tomodel changes in geological features such as sediment prov-enance (Armstrong et al., 2004; Pearce et al., 2005a, b), paleocli-mate (Retallack, 1997; Armstrong et al., 2004; Ratcliffe et al.,2010), basin anoxia (English, 1999; Algeo et al., 2004; Ratcliffeet al., 2012a; Sano et al., 2013), fluctuations in terrigenous input(Davies et al., 2014) or simple lithological variations. However,despite the numerous studies and their proposed geological inter-pretations, there is very little published work (North et al., 2005;Davies et al., 2014) in which chemostratigraphic correlations aretested against physical correlations that are chronostratigraphic.

This paper documents the correlation of fluvial strata usinggeochemical profiling in the Beaufort Group of the Karoo Basin,South Africa, and tests these correlations against absolute strati-graphic correlations derived from physically linking sections inthe field. In doing so, it aims to (1) critically assess the validityand limits of geochemical profiling as a correlation tool in fluvialsuccessions, (2) explore possible explanations of the correlationsin terms of fluvial architecture which would produce the elementalabundances on the temporal and spatial scales found in this study,and (3) consider whether this study is likely to be the exception orthe rule that chemostratigraphy can provide chronostratigraphi-cally significant correlations in fluvial settings.

STUDY AREA AND STRATA

The study sections are located in the Karoo Basin of South Africa,west of the town of Sutherland in the Northern Cape (Figure 1).The sequence being analyzed belongs to the Permian Waterford

G. V. Hildred ∼ Chemostrat Inc., 750 Bering,Suite 550, Houston, Texas 77057;gemmahildred@chemostrat.com

Gemma Hildred graduated from the University ofBirmingham with an M.Sci. Honors degree(2006). Gemma is a divisional manager atChemostrat Inc. (Houston), with her main focusesbeing low-accommodation fluvial sequences inWestern Canada and the chemostratigraphiccharacteristics of shale resources plays, such asthe Eagle Ford Formation in Texas and the HornRiver Formation of British Columbia, Canada.

S. S. Flint ∼ University of Manchester, School ofEarth, Atmospheric and Environmental Sciences,United Kingdom; stephen.flint@manchester.ac.uk

Stephen Flint holds a personal chair instratigraphy and petroleum geology at theUniversity of Manchester (United Kingdom). Afterobtaining his Ph.D. from the University of Leeds(1985), he worked for Shell Research(Netherlands), before setting up the StratigraphyGroup at the University of Liverpool in 1991. In2012, he moved the Stratigraphy Group to TheUniversity of Manchester.

ACKNOWLEDGEMENTS

The Chemostrat and Chevron authors would liketo thank their respective companies for providingtime and support in the preparation of thisdocument. We are all indebted to PaulMontgomery and Jösta Vermeulen, withoutwhom the project from which this paper hasbeen prepared would never have taken place.Further, the authors also wish to thank DougCole, Ajay Mistry, Andy Palfrey, John Kavanagh,Alice Gulliford, Laura Edwards, John Howell, andSimon Buckley. This work was funded byChevron Australia Pty Ltd, and the FORCESedStrat Group provided partial funding of theHeli LIDAR collection.

EDITOR ’S NOTE

A color version of Figure 3 can be seen in theonline version of this paper.

156 Ground Truthing Chemostratigraphic Correlations in Fluvial Systems

and Abrahamskraal Formations of the Ecca andBeaufort Groups, respectively (Figure 2).

The Waterford Formation is a marine sequenceof shallow marine mudstones and sandstones. TheWaterford Formation at Ouberg Pass is character-ized by (1) submeter-thick, grey to dark green lami-nated, massive, rippled or heterolithic (with finesandstone) fine to coarse siltstones with carbonatenodules, and occasional Glossopteris leaf fossils;(2) submeter-thick, massive, laminated and rippledfine-grained sandstone sheets with occasionalSkolithos trace fossils and nodules; and (3) thicker(≤ 12 m [≤ 39 ft]) laminated, rippled, massive andtrough cross-bedded, fine-grained sandstones withbasal matrix-supported conglomerates. The largersand bodies we interpret to reflect delta mouth bardeposition and the mudstones deposition in interdistri-butary bays.

The Waterford Formation passes upward into afluvial sequence of the Abrahamskraal Formation,lower Beaufort Group. These rocks are characterizedby (1) red, green, or grey dm-bedded, massive,

rippled or laminated, fine- to medium-grained silt-stone sheets, featuring Skolithos burrows, plant frag-ments (rare), pedogenic carbonate nodules, anddesert roses (pseudomorphs after gypsum); (2) blue-grey (weathering beige), meter-scale, massive, lami-nated or rippled, coarse siltstone to fine-grained sand-stone sheets with occasional Skolithos; and (3) larger(>4 m [> 13 ft] thick) laminated, low-angle planarcross-bedded or rippled (rarer) fine- to (rare)medium-grained sandstone sheets with occasionalbasal conglomerates, forming thicker complexeslocally (Wilson et al., 2014). The facies proportionsare 0.2–0.9% conglomerates, 17.0–26.8% sandstone,72.0–82.8% mudstone, and 0.1–0.4% tuffs, by thick-ness from Skurweberg and Ouberg logs. These areinterpreted to represent deposition by (1) distal cre-vasse splays and ephemeral lake and flood basin sus-pension settling deposition, (2) proximal crevassesplays, and (3) bed load deposition within fluvialchannels (channel belts), respectively. Desiccationcracks and wind-accumulated gypsum crystals(Stear, 1978) also indicate strong evaporation.

Figure 1. Study area and location of three study sections. (A) Outline map of southern Africa showing the location of box B. (B)Geological map of the western Karoo Basin showing the basin stratigraphy and the location of the study area C. (C) Hill-shaded topo-graphic map of study area highlighting the location of the three logged sections. Lines of alternating white and black indicate the loca-tions of sampled sections, solid black lines are tar roads or gravel tracks. Geological map based on Johnson et al. (1997).

RATCLIFFE ET AL. 157

White, centimeter- to decimeter-thick volcanic tuffsare present within the sections but are apparently rare,although their rarity may be a feature of hillslopeweathering as they are more common in roadcuts than in natural exposure hillside and streamsections.

Three sections have been logged and analyzedfrom Ouberg Pass (OP) (log 0 m, −32.407412°,20.326497°), Skurweberg Pass (SP) (log 0 m,−32.351120°, 20.348638°) and VerlatekloofPass (VP) (log 0 m, −32.562565°, 20.526007°)(Figures 1, 3). The Ouberg Pass section runs fromthe top of the Waterford Formation into the overly-ing Abrahamskraal Formation. The section totals549 m (1801 ft) and exposure is generally good

(Figure 3B). The section at Skurweberg Pass liesapproximately 7 km (4.4 mi) to the north of theOuberg Pass section and is 222 m (728.4 ft) inlength, and exposure along this section is moderateto good (Figure 3A). The base of the VerlatekloofPass Section lies 25.5 km (15.8 mi) to the southeastof the base of the Ouberg Pass. The total thicknessof the section is 739 m (2424.5 ft), with about 50%exposure over this thickness. The sections atOuberg and Skurweberg Passes were loggedthrough hillside exposures, and at VerlatekloofPass several discrete road cut exposures along theR354 road were logged, separated by areas of non-exposure. Exposure in the road cuts is good(Figure 3C), but between the cuts is poor or

Figure 2. Stratigraphy of study area. The study interval, highlighted by a box, is located in the mid-Permian Abrahamskraal Formationand the underlying Waterford Formation. Based on Flint et al. (2011) and Wickens (1994).

158 Ground Truthing Chemostratigraphic Correlations in Fluvial Systems

completely covered. The thickness of unexposedstrata was estimated in the field using a laser ranger(LTI Trupulse 360B) and bedding dips, and verifiedusing Google Earth to trace exposed sandstonedeposits from the inaccessible exposures into theroad cut. All sections were logged at cm–dm resolu-tion where bedrock was exposed, and sampled at50 cm (19.7 in.) intervals for elemental analysisallowing samples to be confidently tied tostratigraphy.

No regional marker surfaces or strata exist, suchas marine limestones, and although local sand-richand mud-rich intervals are present, they are notcorrelative locally or regionally. There are also nosystematic or diagnostic lithostratigraphic featureswhich differentiate any part of Ouberg, Skurweberg,or Verlatekloof sections from any other of the threesections.

STUDY MATERIAL

Field samples were collected and analyzed fortheir whole rock inorganic geochemistry, with theaim of defining a chemostratigraphic zonation forthe sequence. A total of 1075, 276, and 412 sampleswere collected from Ouberg, Skurweberg, andVerlatekloof Passes, respectively. In addition tograin-size data obtained from sedimentary logging,the elemental data have been used to create a “chemi-cal gamma” log. K2O, Th, and U were all acquired bythe processes described below, thereby allowing theAPI value of each sample to be determined using theformula:

Chemical Gamma = ðK2O × 13.55Þ + ðTh × 3.93Þ+ ðU × 8.08Þ

(Ellis and Singer, 2007). When these values areplotted, a log comparable to a gamma (ray) log is pro-duced (Figure 4).

Sample Preparation and Analysis

Where possible, samples that showed no influenceof surficial weathering were collected. Weatheringrinds were carefully removed prior to preparation.

Each sample was described and located on the sedi-mentary logs for Ouberg and Skurweberg passes.They were then ground to a powder in a ball milland, following Li-metaborate fusion, were analyzedusing inductively coupled plasma optical emissionspectrometry (ICP-OES) and mass spectrometry(ICP-MS). The methods of fusion and analysis arethose described in Jarvis and Jarvis (1995). Theseanalytical methods result in data for 50 elements:10 major elements, reported as oxide percent byweight (SiO2, TiO2, Al2O3, Fe2O3, MgO, MnO,CaO, Na2O, K2O and P2O5); 25 trace elements,reported as parts per million by weight (ppm) (Ba,Be, Bi, Co, Cr, Cs, Cu, Ga, Hf, Mo, Nb, Ni, Pb,Rb, Sn, Sr, Ta, Tl, Th, U, V, W, Y, Zn, and Zr);and 14 rare earth elements (REE), reported as partsper million by weight (La, Ce, Pr, Nd, Sm, Eu, Gd,Tb, Ho, Dy, Er, Tm, Yb, and Lu). Precision errorfor the major element data was generally better than2% and is around 3% for the high-abundance traceelement data derived by ICP-OES (Ba, Cr, Sc, Sr,Zn, and Zr). The remaining trace element abundan-ces were determined from the ICP-MS, and datawere generally less precise, with precision error onthe order of 5%. Accuracy error is ± 1% for majorand ± 3 to 7 ppm for trace elements, depending onthe abundance. Expanded uncertainty values (95%confidence), which incorporate all likely errorswithin a statistical framework derived from 11batches of 5 certified reference materials (CRMs),each prepared in duplicate, are typically 5–7% (rela-tive) for major elements and 7–12% (relative) fortrace elements.

CHEMOSTRATIGRAPHICCHARACTERIZATION AND CORRELATION

A chemostratigraphic study aims to identify packagesof strata which have unique chemical signatures thatcan be identified in a stratal sequence by the relativechanges in chemical composition throughout the sec-tion. The primary control on elemental geochemistryof sediments is lithology (rock-forming mineralogy),but chemostratigraphy relies upon characterizationof subtle changes within similar lithologies.Therefore, the first step in a chemostratigraphic

RATCLIFFE ET AL. 159

Figu

re3.

Overviewof

threesampled

sections.(A)

Oblique

aerial

view

ofSkurwebergPass.Selectedstratigraphicheightsareshow

n.(B)View

oftheup

perpo

rtionof

theOub

ergPass

sectionfrom

GoogleEarth

constru

cted

from

plan-viewaerialphotographs

draped

onto

adigitalelevationmodel.(C)

Road

cutexposureform

ingpart

ofVerlatekloofPass.Our

logalongVerlatekloofPassiscomposedof

aseriesofroad

cutsseparatedby

areasofpoor

orno

exposure.

160 Ground Truthing Chemostratigraphic Correlations in Fluvial Systems

characterization is to divide the dataset into broadlysimilar lithological groups. No bias was placed onlithology during sampling, that is, the lithologyencountered at each 50 cm (19.7 in.) increment wasselected irrespective of whether it was a sandstone,siltstone, or mudstone (claystone). As a result, a con-tinuous array of siliciclastic lithologies have beenanalyzed with no clear geochemical demarcationbetween sandstone and mudstone lithologies possible(e.g., by SiO2∕Al2O3 values). This makes defininglithological groups somewhat arbitrary. Based on thefield description of the strata from where the sampleswere taken, the geochemical data related to that sam-ple were divided into one of the following twogroups: (1) medium to coarse siltstones ranging upto fine to medium sandstones and (2) fine siltstonesand mudstones (claystones). These two lithologicalgroups were considered separately for chemostrati-graphic purposes; the medium siltstones and sand-stone collectively were referred to as the coarselithologies, and fine siltstones and mudstones (clay-stones) as the fine lithologies.

The aim of this work is to validate the use ofchemostratigraphy as the basis of correlation fromdata derived from wellbore samples in fluvial reser-voirs, the vast majority of which are cuttings (e.g.,Ratcliffe et al., 2010). Cuttings samples from well-bores are rarely collected at a sampling frequencygreater than 3 m (9.8 ft), and these are compositesamples containing representative components fromthe entire 3 m (9.8 ft) of their depth range. To mimicresults from cuttings, a moving average was appliedto the field sample data, with n = 10, that is, a 5 m(16 ft) composite. This process also has the effectof better enabling trends in the data to be seenthrough background noise. When dealing with com-posited cuttings samples, the predominant lithologyover the composited interval is picked from the cut-tings sample for analysis. Because regional correla-tions commonly rely upon mudstones (e.g., Pearceet al., 2005a; Ratcliffe et al., 2010), mudstones werepreferentially selected. As a consequence, sandstonebeds less than 1–2 m (3.3–6.6 ft) thick (dependingon sample frequency) were rarely sampled from thecuttings samples. Therefore, only data from beds ofcoarse lithology over 2 m (6.6 ft) in thickness aredisplayed in the chemical logs of Figure 4.

Chemostratigraphic characterization, like anystratigraphic technique, requires a hierarchicalapproach with definition of broad isochemical inter-vals, which can typically be defined using graphicalmethods, and thinner intervals that may be basedmore on chemical log character than absolute changesin composition. The nomenclature and protocol ofRatcliffe et al. (2010) are followed here, such thatthe broad graphically recognizable intervals arereferred to as packages and the finer interval as geo-chemical units. Typically, the chemostratigraphicpackages are used to define correlations over tens ofkilometers, and geochemical units are used to definehigher resolution correlations over several kilometers(Pearce et al., 2010; Ratcliffe et al., 2010, 2012b;Sano et al., 2013).

Fine Lithology Geochemistry

The entire study interval is divided into eight graphi-cally distinguishable packages, package 10 to pack-age 80, based largely on characterization of the finelithologies (Figures 4, 5). Although not all packageswere recognized in all sections, because of the lim-ited exposure rather than original stratigraphy, somecan be correlated between all three sections(Figure 4). Between the two relatively closelyspaced sections at Ouberg Pass and Skurweberg Pass,several geochemical units were recognized and corre-lated (Figure 6).

Packages 10–30

These three packages are restricted to the OubergPass Section (Figure 7) and represent the transitionfrom marine deltaic sediments of the WaterfordFormation (package 10) to fluvial sediments of theBeaufort Group (package 30 and above).

• Package 10: The fine-grained lithologies in thispackage have higher TiO2∕Nb values than anyothers in the study interval (Figures 4–6).

• Package 20: The fine-grained lithologies in thispackage have lower TiO2∕Nb values than under-lying package and higher Cr/Sc and Cr/Nb valuesthan the overlying package (Figures 4–6).

RATCLIFFE ET AL. 161

Figu

re4.

Chem

ostratigraphiccorrelationbetweenOubergPass,SkurwebergPass,and

VerlatekloofPassw

ithchem

icallogsconstructedfro

mfine-grainedlithologies

(graysquares)

andfro

mcoarse-grained

lithologies

(yellowsquares).A

splined

movingaveragehasb

eenappliedtothedatatobettersee

trendsthrough

background

noise

andtomimica5m(16ft)

composited

cuttingssamplefro

mawellbore.SeeFigure

1forsectionlocations.

162 Ground Truthing Chemostratigraphic Correlations in Fluvial Systems

• Package 30: The fine-grained lithologies fromthis package are differentiated from those belowby their low TiO2∕Nb and Cr/Sc values and fromthose above by their high TiO2∕Nb values(Figures 4–6).

Package 40

The fine-grained lithologies from package 40 aredifferentiated from those below by their lowTiO2∕Nb values and from those above by their low

Cr/

Sc

Cr/

Sc

Cr/

Sc

TiO2/Nb TiO2/Nb

TiO2/Nb

Package 70Package 60Package 50Package 40Package 30Package 20Package 10

Package 80Package 70Package 60Package 50Package 40

Ouberg Pass Verlatekloof Pass

3.0

3.2

3.4

3.6

3.8

4.0

4.2

4.4

4.6

4.8

5.0

3.0

3.2

3.4

3.6

3.8

4.0

4.2

4.4

4.6

4.8

5.0

0.040 0.045 0.050 0.055 0.060 0.040 0.045 0.050 0.055 0.060

0.040 0.045 0.050 0.055 0.0603.0

3.2

3.4

3.6

3.8

4.0

4.2

4.4

4.6

4.8

5.0

Package 70

Package 70

Package 70

Package 60

Package 60

Package 60

Package 50

Package 50

Package 50

Package 10

Package 10

Packages 20 and 30Package 40

Package 40

Package 40

Package 80

Package 80

Packages 20 and 30

Ouberg Pass

Ouberg and Verlatekloof

Verlatekloof Pass

Figure 5. Binary diagrams constructed from fine-grained lithology data to graphically display package definition in Ouberg Pass andVerlatekloof Pass. Note that a splined moving average has been applied to the data to better see trends through background noise.

RATCLIFFE ET AL. 163

Sku

rweb

erg

Pas

s K2O

/Na 2O

Coa

rse

100

0 -100

clay packages

Package 70 Package 60 Package 50

Na

2OA

l 2O3

Coa

rse

Tops

60.4

60.3

60.2

60.1

50.2

50.1

Ou

ber

g P

ass

500

400

300

Che

mic

alG

am

ma

Che

mic

alG

am

ma

clay packages

Package 70 Package 60 Package 50 Package 40

Tops

60.4

60.3

60.2

60.1

50.2

50.1

Cr/

TiO

2

Coa

rse

K2O

/Na 2O

Coa

rse

Na

2OA

l 2O3

Coa

rse

Cr/

TiO

2

Coa

rse

TiO

2 / N

bF

ine

Cr

/ Sc

Fin

eC

r / N

bF

ine

Rb/

Cs

Fin

eT

iO2

/ Nb

Fin

eC

r / S

cF

ine

Cr

/ Nb

Fin

eR

b/C

sF

ine

Hei

ght

m

Hei

ght

m

Figu

re6.

High-re

solutionchem

ostratigraphiccorrelationbetweenOubergPassandSkurwebergPasswith

chem

icallogsconstructedfro

mfine-grainedlithologies(graysquares)and

fromcoarse-grained

lithologies

(yellowsquares).A

movingaveragehasbeen

appliedtothedatatobettersee

trendsthroughbackground

noise.See

Figure

1forsectionlocations.

164 Ground Truthing Chemostratigraphic Correlations in Fluvial Systems

Cr/Sc and Cr/Nb values. The top of Package 40 isdefined by a sharp upward increase in Cr/Sc andCr/Nb values and marks one of the most significantgeochemical changes in the study sequence, a sur-face essentially dividing the study interval into two(Figures 4–6).

Package 50

The fine-grained lithologies of this package aredifferentiated from those below by their high Cr/Scand Cr/Nb values and from those above by their highTiO2∕Nb values (Figures 4–6).

3.0

3.2

3.4

3.6

3.8

4.0

4.2

4.4

4.6

4.8

5.0

0.05 0.05 0.06 0.063.0

3.2

3.4

3.6

3.8

4.0

4.2

4.4

4.6

4.8

5.0

0.05 0.05 0.06 0.06

3.0

3.2

3.4

3.6

3.8

4.0

4.2

4.4

4.6

4.8

5.0

0.05 0.05 0.06 0.06

Cr/

Sc

Cr/

Sc

TiO2/Nb TiO2/Nb

TiO2/Nb

Cr/

Sc

Package 70Package 60Package 50Package 40Package 30Package 20Package 10

Package 70Package 60Package 50

Ouberg Pass Skurwerberg Pass

Ouberg Pass

Ouberg and Skurweberg

Skurweberg Pass

Package 10

Package 10

Package 60

Package 60

Package 60Package 50

Package 50

Package 50

Package 70

Package 70

Package 70

Package 40

Package 40

Packages 20 & 30

Packages 20 & 30

Figure 7. Binary diagrams constructed from fine-grained lithology data to graphically display package definition in Ouberg Pass andSkurweberg Pass. Note that a moving average has been applied to the data to better see trends through background noise.

RATCLIFFE ET AL. 165

Package 60

The fine-grained lithologies of this packageare differentiated from those in the underlyingpackage by their low TiO2∕Nb values and fromthose above by their low Cr/Sc and Cr/Nb values(Figures 4–6).

Package 70

The fine-grained lithologies of this packageare differentiated from those in the underlyingpackages by their high Cr/Sc and Cr/Nb values(Figures 4–6).

Package 80

The fine-grained lithologies of this package aredifferentiated from those in the underlying packagesby their low TiO2∕Nb values (Figures 4–6).

Coarse Lithology Geochemistry

The coarse-grained lithologies also show significantchanges in elemental composition through time, withthree clearly defined long-lived compositions. Afourth composition is seen, but appears to be strati-graphically restricted (Figure 8).

Composition 1 is seen in the Ouberg Pass Sectionin the coarser lithologies from Package 10. Values ofNa2O∕Al2O3 and Ga/Rb are low in this compositioncompared to those above.

Composition 2 is seen on the OubergPass Section and at the base of the Verlatekloof PassSection. The coarse lithologies that represent thiscomposition have higher Na2O∕Al2O3 and Ga/Rbvalues than those of composition 1 and lowerNa2O∕Al2O3 values than those above. The values ofGa/Rb reach a maximum at the sandstone midwaythrough Package 40. In both the Ouberg Pass andVerlatekloof Pass sections, the change in coarse-grained lithology composition is abrupt and occurs

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

10 30 50 70 90

TiO

2 (%

)

Cr (ppm)

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.15 0.20 0.25 0.30 0.35 0.40 0.45

Ga/

Rb

Na2O/Al2O3

Comp 3Comp 2Comp 1

Comp 4Comp 3

Comp 4Comp 3Comp 2

Ouberg VerlatekloofSkurweberg

Composition4

Composition2

Composition3

Composition 1

Compositions1-3

Figure 8. Binary diagrams constructed from coarse-grained lithologies, showing the four sandstone elemental compositions encoun-tered in the study intervals. Refer to Figure 4 for vertical distributions of compositions in study sections.

166 Ground Truthing Chemostratigraphic Correlations in Fluvial Systems

around 25–30 m (82–98 ft) above the base of Package40 (Figure 4), or the first significant coarse lithologyabove the top of Package 40.

Composition 3 is seen in all three sections, and thecoarse-grained lithologies of this composition have thehighest Na2O∕Al2O3 values in the study interval.

Composition 4 is restricted to sandstones atthe base of Package 70. This composition has veryhigh Cr values, resulting in high Cr∕TiO2 values(Figures 4, 8).

CHEMOSTRATIGRAPHIC CORRELATIONS

Vertical changes in elemental composition can clearlybe recognized in the study sections (Figure 4).However, as emphasized by Schoch (1989) correlationis of overriding importance to stratigraphy, and thescience would be reduced to the description ofinnumerable, localized successions of strata withoutit. Characterizing single wellbores is of little impor-tance, but being able to extend that characterizationover an oil field or a prospective basin is without doubtthe single most important building block for alllater studies in the oil and gas industry. In this section,the correlation of the three studied sections isconsidered.

Short Range: Ouberg Pass to SkurwebergPass (Field-Scale Correlation)

The fine lithologies from the Skurweberg Passsection plot in three distinct clusters that allow thepackages defined in the Ouberg Pass section to berecognized in the Skurweberg Pass section(Figure 7). The top of package 50 is placed at a sharpupward decrease in TiO2∕Nb values and the top ofpackage 60 at a sharp upward increase in Cr/Sc andCr/Nb values (Figure 6).

In addition to this package correlation, package60 is further divided into four geochemical units 30–50 m (98–164 ft) thick that are correlative betweenthe Ouberg and Skurweberg Pass sections(Figure 6). Unit 60_1 has high TiO2∕Nb values,appearing to be transitional between package 50 andthe overlying units of package 60. Unit 60_2 is char-acterized by its low Rb/Cs values compared to theother units in package 60. Unit 60_3 has higher

Rb/Cs values than Unit 60_2 and lower TiO2∕Nb val-ues than Unit 60_1. Unit 60_4, which has highTiO2∕Nb values, is only seen in the Ouberg PassSection, presumably being associated with a zone ofno exposure in the Skurweberg Pass section.

Long Range: Ouberg Pass to VerlatekloofPass (Subregional-Scale Correlation)

The oldest section in Verlatekloof Pass hasfine-grained lithologies with low Cr/Sc and Cr/Nbvalues that plot in association with package 40samples from the Ouberg Pass section (Figures 5,6). A sharp upward increase in values of the Cr/Scand Cr/Nb ratios is seen 30 m (98 ft) above the baseof the section, which is correlated to the samefeature in the Ouberg Pass section to define the topof package 40. This interpretation is supported bythe change in composition of the coarse lithologiesthat occurs about 30 m (98 ft) above this surfacein both the Verlatekloof and Ouberg sections.(Figures 6, 8).

A sharp upward decrease in TiO2∕Nb values ofthe fine-grained lithologies allows recognition of thetop of package 50 in both sections (Figure 4). Theassignment of samples to package 50 is supportedby the binary diagrams in Figure 5.

Samples of fine-grained lithologies from package60 have distinctively low TiO2∕Nb values and highCr/Sc values in both sections (Figures 4, 5). Althoughthe top of the package is close to the top of the studyinterval in the Ouberg Pass Section, it is relatively welldefined by the upward increase in Cr/Sc and Cr/Nbvalues. In Verlatekloof Pass, this change appears tooccur in a zone of no exposure.

The coarse lithologies at the base of package 70in the Verlatekloof Pass Section have very highCr∕TiO2 values, a similar composition to that seenat the base of this package in the Skurweberg PassSection; no coarse lithologies were analyzed fromPackage 70 in the Ouberg Pass Section.

The Verlatekloof section extends almost 450 m(1476 ft) stratigraphically above the top of theOuberg Pass Section. One notable and consistentchange in geochemistry is recognized in this intervalat 460 m (1509 ft) above section base, where the

RATCLIFFE ET AL. 167

TiO2∕Nb values decrease. The low TiO2∕Nb valuesare used to define package 80 (Figures 4, 5).

Interpretation of Key Elements and ElementRatios Used to Define the ChemostratigraphicCharacterization of the Beaufort Group

The correlations in Figures 4 and 6 were achieved byidentifying trends in the ratios of selected elementconcentrations. Here, the meaning of those ratios isdiscussed. Ideally, a wide suite of mineralogicaldata, petrography, heavy mineral analysis, andX-ray diffraction data would be required to

comprehensively understand the mineralogical con-trols on the elements being used for stratigraphiccorrelations. Such data are unavailable here and,indeed, are rarely available in an oilfield setting.Numerous authors (Pearce et al., 2005a; Svendsenet al., 2007; Ellwood et al., 2008; Pe-Piper et al.,2008; Ratcliffe et al., 2010, Sano et al., 2013) havedemonstrated that an indication of the main mineral-ogical controls on elemental data can be gained byconsidering the eigenvector values obtained throughprinciple component analysis (PCA). Eigenvectors 1and 2 for the fine lithologies and for the coarselithologies are presented in Figure 9. In this type of

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

-0.1 0 0.1 0.2 0.3

Al2O3

SiO2

TiO2

Fe2O3

MnO

MgO

CaO

Na2O

K2O

P2O5

Ba

Cr

Sc

Sr

Zn

Zr

V

Co Ni

Cu

Ga

Rb

Y

Nb

Mo

Sn

Cs

LREE

MREE

LREE

Hf

Ta Th

U

-0.2

-0.1

0

0.1

0.2

0.3

0.4

-0.2 -0.1 0 0.1 0.2

Al2O3

SiO2

TiO2

Fe2O3

MnO

MgO

CaO Na2O

K2O

P2O5

Ba

Cr Sc

Sr

Zn

Zr

Be

V

Co

Ni

Cu

Ga

Rb

Y

Nb

Mo

Sn

Cs

MREE

HREE

LREE Hf

Ta Th

U

Eigenvector 1

Eigenvector 1

Eig

enve

ctor

2E

igen

vect

or 2

Group 1

Group 1

Group 2

Group 2

Group 3

Group 3 Group 4

Group 5Group 6

Group 4

(A)

(B)

Figure 9. Principal componentanalysis. Eigenvector 1 versus eigen-vector 2 crossplots. (A) is constructedfrom data derived from fine-grainedlithologies, and (B) is constructed fromdata derived from coarse-grainedlithologies. Eigenvectors are derivedfrom principal components analysis(PCA). LREE = light rare earth elements(La, Ce, Pr, Nd, Sm), MREE = middlerare earth elements (Eu, Gd, Tb, Dy),and HREE = heavy rare earth elements(Ho, Er, Tm, Yb, and Lu).

168 Ground Truthing Chemostratigraphic Correlations in Fluvial Systems

diagram, the more closely elements plot to oneanother, the more closely they are related to oneanother in the sediment, so two elements plotting inextremely close proximity (e.g., Zr and Hf onFigure 9) are likely to be present in the same mineralphase, which in the case of Zr and Hf is the heavymineral zircon. In Figure 9, the elements fall intowell-defined groups that allow the interpretationbelow to be made.

Mineralogy: Fine-Grained Lithologies

The eigenvectors for elements of the fine-grained lith-ologies (Figure 9A) fall into six groups.

Group 1 contains SiO2 and Na2O. In silty mud-stone lithologies, the amount of SiO2 reflects theamount of quartz. Therefore, the close association ofNa2O with SiO2 suggests that it is in some mannerassociated with quartz in the fine-grained lithologies.The most common way this association develops isbecause of the grain-size similarity of plagioclasefeldspar and quartz, implying that Na2O in the fine-grained lithologies of the study interval is associatedwith plagioclase feldspar. A similar association isreported by Ratcliffe et al. (2010) for silty mudstonesin the Mungaroo Formation, Northern CarnarvonBasin, Australia.

Group 2 contains Al2O3, K2O, MgO, Fe2O3, Ga,Zn, Sc, Ni, Cr, Rb, Co, and Cs. In siliciclastic sedi-ments Al2O3 is typically directly related to total claycontent (Pearce et al., 2005a; Ratcliffe et al., 2012b;Sano et al., 2013) and, therefore, the elements associ-ated with Al2O3 are also likely to be associated withclay minerals. K2O, Rb, and Cs are typically associ-ated with illite/smectite, Al2O3 and Ga with kaolinite,and MgO, Fe2O3, Sc, Zn, and Ni with chlorite.

Group 3 contains Zr and Hf, both of whichare typically associated with zircons in siliciclasticsediments (Armstrong et al., 2004; Pe-Piper et al.,2008).

Group 4 contains P2O5, light rare earth elements(LREE), middle rare earth elements (MREE), heavyrare earth elements (HREE), and yttrium (Y).Discrete tuff beds occur in the study intervals, andthese have high concentrations of the elements in thisgroup. Although data from tuffs are not included inthe dataset used to generate the eigenvector plots in

Figure 9, these elements probably reflect tufaceousmaterial entrained in the sediment.

Group 5 contains TiO2, Th, Ta, and Nb. This is acommon group of elements that typically indicatesthe influence of Ti-oxide minerals such as rutile andanatase (Pearce et al., 2010; Ratcliffe et al., 2010).

Group 6 contains CaO, MnO, Sr, Mo, Cu, andBa. CaO is typically associated with calcite and,therefore, the other elements in this group are alsolikely to be associated with calcite.

Based on the discussions above, the followinginferences can be made about the element ratiosshown in Figures 4 and 5.

For Cr/Sc, both elements plot closely together ingroup 2 in Figure 9A, implying that both are associ-ated with clay minerals. Although both are typicallymost abundant in chlorite, the changes in the Cr/Scratio must be recording changes in the proportion ofclay mineral species in the fine-grained lithologies.

For Cr/Nb, Nb plots in group 5, interpreted to berelated to Ti-oxide heavy minerals. Therefore, thisratio most likely reflects changes in the abundanceof chlorite and Ti-oxide heavy minerals.

For TiO2∕Nb, both elements plot in group 5, andtherefore, changes in this ratio probably reflectchanges in species of Ti-oxide minerals.

Mineralogy: Coarse-Grained Lithologies

In Figure 9B, which is constructed from coarse-grained lithologies, the elements fall into three dis-tinctive groups.

Group 1 contains SiO2 and Na2O. SiO2 typicallyreflects the amount of quartz in siliciclastic sediment,and therefore, Na2O must be associated with quartz.A similar association is reported in Ratcliffe et al.(2010) where the Na2O association is demonstratedto be caused by the amount of plagioclase feldspar.

Group 2 contains Al2O3, K2O, MgO, Fe2O3, Rb,Cs, Ga, Ni, Co, Sc, and Zn. In a siliciclastic sediment,Al2O3 typically reflects total clay contents in thematrix, and therefore, the elements associated withAl2O3 are likely to be associated with clay minerals.K2O, Rb, and Cs are typically associated with illite/smectite, Al2O3 and Ga with kaolinite, and MgO,Fe2O3, Sc, Zn, and Ni with chlorite.

RATCLIFFE ET AL. 169

Group 3 contains TiO2, P2O5, Nb, Ta, Th, Y, Cr,Zr, Hf, LREE (including La), MREE, and HREE. Zrand Hf are typically associated with the amount ofzircon in a sandstone, and therefore, other elementsassociated with Zr and Hf are likely associated withheavy minerals (Armstrong et al., 2004). Typically,P2O5 and La are associated with apatite, Cr with Cr-spinel, and TiO2, Nb, Ta, and Th with Ti-oxide heavyminerals.

Group 4 contains CaO, MnO, Sr, Mo, Cu, andBa. CaO is typically associated with calcite, andtherefore, the other elements in this group are alsolikely to be associated with calcite.

Therefore, the probable controls on the key ratiosused to characterize the coarse lithologies are asfollows.

For Na2O∕Al2O3, Na2O is associated with pla-gioclase feldspar. By normalizing it against Al2O3,some of the minor variations in lithology areremoved. A similar interpretation of Na2O∕Al2O3

was made by Ratcliffe et al. (2010) in the sandstonesof the Mungaroo Formation.

For Ga/Rb, both elements are associated withclay minerals, but Ga is more commonly associatedwith kaolinite and Rb with illite. Therefore, this ratioreflects the proportion of kaolinite and illite. A simi-lar interpretation of Ga/Rb was made by Ratcliffeet al. (2010) in the sandstones of the MungarooFormation.

For Cr∕TiO2, the high value of this ratio inFigure 4 is driven by Cr and the high Cr∕TiO2 sandsare interpreted to have a distinctive mafic provenance.

PHYSICAL CORRELATIONS

The chemostratigraphic correlations were testedagainst physical correlations established usinghelicopter-based LIDAR and Google Earth, for theshort and long distance correlations, respectively.

Short Range: Helicopter-Based LIDARCorrelation

The hillside section between Skurweberg and Oubergpasses, 7 km (4.4 mi) apart, were scanned using ahelicopter-mounted laser scanner with integrated

high-resolution digital camera to collect a point cloudrepresenting the topography and photographs of thesection. These were combined into a photo-realisticvirtual outcrop model (VOM) using the processingtechniques outlined in Buckley et al. (2008) andRittersbacher et al. (in press). Channel belt architec-ture in the area is characterized by sheet-like channelbelt geometries with semi-continuous, horizontalupper bounding surfaces, except for the locations ofabandoned mud-filled channels (mud plugs) whichare, however, rare (architectural style 1 of Wilsonet al., 2014). The sandy channel belt facies aredemonstrably discontinuous, even between theOuberg and Skurweberg sections, with an average ap-parent width of channel belts of 550 m (1804.5 ft),derived from LIDAR data in the area. Sand bodiesand siltstone sheets were walked out in the field.The LIDAR data shows that no channel belts extendbetween the two sections.

The physical correlation was achieved by tracingout the tops of three geographically overlappingchannel belts on the VOM between Ouberg andSkurweberg Passes (Figure 10). The stratigraphicthickness between the channel belt tops (jumps inFigure 10) were measured on the VOM, and the netstratigraphic thickness between the points on adjacentlogs was calculated. This process matched a knownposition on the Skurweberg Pass log to a known posi-tion on the Ouberg Pass log, establishing a robustphysical correlation between the two logs (Figure 10).The result is a correlation where the 126 m (413.4 ft)position on the Skurweberg log is stratigraphicallyequivalent to the 498 m (1634 ft) position on theOuberg Log (Figure 10). We estimate that this methodis accurate to within ± 5 m (± 16 ft). Although thelogged sections were 222 m (728 ft) and 549 m(1801 ft) thick, for Skurweberg Pass and OubergPass, respectively, only the upper 100 m (328 ft) ofOuberg Pass was captured in the VOM. The physicalcorrelation, therefore, is a single link between onestratigraphic marker bed on Skurweberg Pass and asecond on Ouberg Pass. A kilometer-scale depositio-nal paleosurface was discovered on the top of a silt-stone sheet (Figure 3). Depositional paleosurfaces areancient surfaces, commonly containing trace fossilsor in situ plant fossils, which have been lithified andburied (see Smith, 1998). The entirety of such surfaces

170 Ground Truthing Chemostratigraphic Correlations in Fluvial Systems

Figu

re10

.LIDA

R-basedphysicalcorrelationbetweenSkurwebergPassandOubergPass.Fourp

anels(A–D)

show

images

betweenthetwosections

ofthetopographicmodelonly.

Sandsto

nechannelbeltsareexpressedon

thehillslope

assm

allcliffs(whiteinimages).Correlationisachieved

bytracin

gthetopofthreechannelbeltsbetweenthetwosections

and

measuringthestratigraphicthickness(jumps)b

etweenthechosen

surfaces.Blackandwhitedashed

lines

arethelogs,black

dashed

lines

indicatethetops

ofthethreechannelbelts

tracedout.Symbolsindicatingequivalent

pointson

adjacentimages

areinclu

dedtoassistthe

reader

intracin

galongthepanels.

RATCLIFFE ET AL. 171

were exposed, whether subaerially or subaqeuously, atthe same time point in the past. Based on the kilo-meter-scale depositional paleosurface at SkurwebergPass (Figure 3), we infer a very low-relief paleoland-scape and, subsequently, that the alignment of thetwo logs to this physical correlation results in adjacentpositions on two logs having the equivalent age, thatis, they are chronostratigraphic, although the specificfacies may vary.

The physical correlation achieved using LIDARdata between Skurweberg Pass and Ouberg Pass veryclosely matches the correlation achieved using che-mostratigraphy (Figure 6). It parallels the package50–package 60 boundary. The minor difference inthe location of the package 60–package 70 boundaryis because this is based on fine lithology chemostra-tigraphy, which is absent in Skurweberg Pass as aresult of the presence of a channel belt at this loca-tion. Over short distances, here 7 km (4.4 mi), chemo-stratigraphy provides a very similar correlation to thephysical correlation but has the advantages ofallowing much higher resolution subdivision of thestratigraphy (Figure 6). This could not be achievedusing the LIDAR data alone, as correlation using theVOM relied exclusively on sandstone outcropswhose widths were shorter than the distance between

the logged sections. Mudstone exposures were onlypresent along the road cut sections or in gullies andwere not visible in the VOMs, so could not be usedto physically correlate the sections.

Long Range: Google Earth Correlation

Verlatekloof Pass is approximately 25.5 km (15.8 mi)southeast of Ouberg Pass. The hillsides between thesesections were not captured by our airborne LIDARsurvey. Only a thin veneer of colluvium exists onthe escarpment hillsides resulting in semicontinuousexposure of strata between the sections. This can beviewed on Google Earth imagery, which has sub-meter scale resolution in this area. Google Earth wasused to trace laterally from the 0 m position on ourVerlatekloof Pass log to the correlated position onOuberg Pass. The position of every 5 m (16 ft) inthe stratigraphic logs was recorded by global posi-tioning system (GPS) during logging, which allowedthe stratigraphic heights of the correlation points tobe determined from their physical locations. Tracingof strata between the two sections was achieved,resulting in the correlation of the 0 m position onour Verlatekloof Log to the 220 m (721.8 ft) positionon the Ouberg Log (Figure 11). The dominant source

Figure 11. Image of the physicalcorrelation made on Google Earthbetween Ouberg Pass and VerlatekloofPass logs. The correlation was achievedby tracing out the strata between thetwo sections. The total length of thetrace is 70 km (43.5 mi) with a distancebetween the sections of 25.5 km (15.8mi).

172 Ground Truthing Chemostratigraphic Correlations in Fluvial Systems

of error in this correlation technique is the difficultyresolving the thickness of channel belt cliff faces.Error in the correlations achieved using this methodis greater than that achieved using LIDAR data andis estimated to be on the order of ± 30 m (± 98 ft).

Over the 25.5 km (15.8 mi) distance betweenOuberg Pass and Verlatekloof Pass, the chemostrati-graphic correlation matches the physical correlationbased on Google Earth mapping. High-resolutionsubdivision of the Ouberg Pass and VerlatekloofPass sections was hindered by lack of exposure, butlarge-scale (e.g., >30 m [> 98 ft]) trends and uniquefeatures, such as the sandstone composition 4 (highCr∕TiO2), could be identified that provided moreuseful data than provided by lithostratigraphy alone.

DISCUSSION

The primary aim of this paper is an assessment of theuse of chemostratigraphy for correlation purposes influvial systems. The data presented above show thatsystematic trends occur in element ratios in a succes-sion of fluvial strata and that these are laterally exten-sive, allowing robust correlations to be made.Furthermore, these correlations mimic physical chro-nostratigraphic correlations over short (about 7 km[4.4 mi]) and moderate (about 25.5 km [15.8 mi]) dis-tances, which in the example used herein, chemostra-tigraphic methods provide chronostratigraphiccorrelations.

Although the study here demonstrates that che-mostratigraphy provides a reliable means of correlat-ing chronostratigraphic surfaces in the fluvialsediments of the Beaufort Group, a broader questionis not only why it was successful in this study, butwhat limitations there may be to its success in otherfluvial systems.

Controls on Geochemical Composition ofFluvial Strata

As concentration is a measure of mass per unit volume,it is obvious that the minerals contained within a sedi-ment sample and the volumetric proportions of thoseminerals are the primary controls on the geochemistry.The specific mineral controls on the study intervals of

this paper are discussed above. However, more generi-cally, it is apparent that many physical and chemicalprocesses affect the primary depositional mineralogy,and changes in the bulk mineralogy can occur by sec-ondary addition of minerals (e.g., by cements and coat-ings) or replacement of minerals by diageneticprocesses (e.g., neomorphism). Other processes canalso affect the volumetric proportion of mineralspresent in a deposit, for example, through processesaffecting the amount of mineral grains present asopposed to which minerals are present (e.g., differenc-es in resistance to abrasion by mineralogy, coupledwith selective sorting) and processes affecting the vol-ume those mineral grains occupy (e.g., compaction). Itis impossible in a practical sense, therefore, to explainall the processes responsible for the elemental compo-sition of a sediment package because of the large num-ber of processes that could have operated to affect itsmineralogy and the volumetric proportion of thoseminerals from deposition to exhumation and sampling.

The aim of any chemostratigraphic study, how-ever, is not to unravel all the processes. Instead, it isto determine whether the subset of elements andelement ratios used for correlation are primarilyreflecting the detrital mineralogy, and if so, whataspects (provenance, clay mineralogy, etc.). Featuresthat likely reflect diagenesis in the elemental data setsuch as CaO (= CaCO3) can be left to one side whenlooking for stratigraphically significant features in anICP-derived elemental dataset. As argued above, wepropose the controls on the elements and elementalratios used to define the correlations displayed inFigures 4 and 6 are changes in detrital mineralogy(plagioclase feldspar, Ti-oxide heavy minerals, andother non-specific mafic minerals) and syndeposi-tional weathering (kaolinite versus illite).

The most likely reason why chemostratigraphiccorrelations in this study provide chronostratigraph-ically significant correlations is because there werechanges in the composition of the sediment beingsupplied to the system (i.e., changes in provenance)and long-term changes in the syndepositional weath-ering style (i.e., paleoclimate), and both of thesechanging parameters are reflected in changes inelemental compositions. These changes irrefutablyrecorded in the sediment are at least approximatelycoeval over a distance of 25.5 km (15.8 mi) as

RATCLIFFE ET AL. 173

demonstrated by the physical correlations usingLIDAR and Google Earth.

Both provenance changes and paleoclimatechanges are the norm, rather than the exception, influvial systems and have been described from a widevariety of settings using a variety of techniques, suchas petrographic analysis and heavy mineral analysis(e.g., Haughton et al., 1991; Morton et al., 2010;Thomas, 2011), paleosol analysis and sedimentology(e.g., Demko et al., 2004), and geochemistry (e.g.,Torres and Gaines, 2013; Van de Velde et al.,2013). Therefore, it has to be assumed that suchchanges are not the limiting factor on the efficacy ofchemostratigraphy in fluvial settings. Instead, it isthe processes which transport and lay down fluvialdeposits and the resulting sedimentary architecture(which varies in size by several orders of magnitude)that will impose limitations on the success of chemo-stratigraphy when it is based on recognizing regionalchanges in provenance and paleoclimate.

Architectural Influences onChemostratigraphy

The scale of the architecture and sedimentary proc-esses involved in transporting and laying down flu-vial deposits must be considered in explainingcorrelations between adjacent geochemical profilesbecause all components of fluvial systems, for exam-ple, channel belts (e.g., Wilson et al., 2014), overbankdeposits (e.g., Smith, 1990), and the fluvial fan meg-astructure (e.g., Nichols and Fisher, 2007) are finitein length, width and thickness, and layer-cake stratig-raphy cannot be assumed over long distances.

Channel belts are the sediment bodies formed bybed load deposited in the base of channels. They arebuilt of smaller sediment packages (e.g., lateral accre-tion packages) such that although two of these sedi-ment packages may be deposited within the channelbelt at roughly the same paleo-elevation, they willdiffer in age, and their mineralogical composition willreflect the composition of sediment that arrived at thatposition at different times. Thus, different positionswithin the channel belt and along its upper surfaceare diachronous, and this will be reflected as noisein the chemostratigraphic profile because of the sam-pling line taken.

Overbank deposits are also limited in lateralextent, although this length scale is greater in lateralextent than channel belts in the study area. For exam-ple, sheets of siltstones, which we interpret to be theproduct of crevasse splaying processes (Wilson et al.,2014), are prevalent in the mudstone deposits of theBeaufort Group and are up to a few kilometers in lat-eral extent. Suspension settling deposits in ephemerallakes can be only as large as the topographichollows between alluvial ridges on the alluvial plain.In fluvial sequences dominated by overbank facies,such as is the case in the Abrahamskraal Formation, itis the fine-grained lithologies that have the greatestpotential to define regionally robust correlationschemes because their widths are more laterally exten-sive than channel belt widths.

Together, channel belt and overbank depositsmake up the alluvial ridge, and repeated avulsion ofthe river channel followed by deposition of anotheralluvial ridge builds up a fan-shape body (megafan)by compensational stacking (e.g., Mohindra et al.1992; Singh et al. 1993; Stanistreet and McCarthy,1993; Gupta, 1997; DeCelles and Cavazza, 1999;Horton and DeCelles, 2001; North et al., 2007;Weissman et al., 2010; Buehler et al., 2011; Fontanaet al., 2013). Based on an analysis of facies andalluvial architecture, Wilson et al. (2014) argued thatmegafans are a good explanation for theAbrahamskraal Formation stratigraphy. At the scaleof the fluvial megafan, only one alluvial ridge isdeposited at any one time, forming a fan lobe(Nichols and Fisher, 2007). The age differencebetween two of the intra-channel belt-scale depositsis small, for example, deposition of lateral accretiondeposits can occur annually, whereas the age differ-ence between two channel belts juxtaposed by avul-sion is much greater because of the relatively lowerfrequency of avulsion. Thus, two adjacent channelbelts in a succession, unless forming part of an ana-branching system (compare to North et al., 2007),may not contain sediment of closely similar deposi-tional ages. And, the primary mineralogical composi-tion of their sediments may have varied, for example,reflecting the evolution of the source catchmentupstream. The main implication of this scale depend-ence of fluvial systems for correlation purposes is thatdeposits of a particular composition of sediment are

174 Ground Truthing Chemostratigraphic Correlations in Fluvial Systems

not deposited across the whole basin in a single, uni-form sheet but in local, lobate deposits.

At the scale of the alluvial ridge, this point isillustrated by adapting the very simple model of allu-vial stratigraphy from Bridge and Leeder (1979). Asediment source with an elemental composition (x/y)varying linearly with time (Figure 12) is applied,where x and y are arbitrary elements. Three logsthrough the model stratigraphy show that the

following features could be expected in chemostratig-raphy of alluvial successions as a result of sedimen-tary architecture at this scale: (1) the overall broadtrend of changing x/y ratio with time is preserved,(2) the stratigraphic record is incomplete in all logs,and (3) strata for the same time intervals are repre-sented by different thicknesses of strata in adjacentlogs, caused by changes in sedimentation rates fromchannel belt to overbank locations. Though this

Figure 12. Model illustrating the variation in completeness and stratigraphic thickness captured by logs through alluvial stratigraphyfor a constant sediment source of linearly varying geochemical composition. (A) Redrawn, a small part of the simple model of alluvialstratigraphy by Bridge and Leeder (1979), which is the basis for the analysis. Channel belts are grey blocks and lines are floodplain sur-faces at the time of avulsion. (B) Bridge and Leeder (1979) model color-coded according to chemical composition (which linearly varieswith time). (C) Vertically exaggerated copy of (B) with position of three logs A–C plotted. (D) Synthetic chemostratigraphic logs throughthe model stratigraphy. (i) The sediment input model consists of a constant sediment source whose ratio of element x to element ychanges linearly with time such that each unit time interval has a composition from 1 to 12. (ii) Compositions representing time intervals5, 7, and 10 are not recorded in Log A because of the compensational architecture of the sequence. Channel belts have higher rates ofdeposition than overbank environments (which is required to build alluvial ridges) resulting in compositions 4 and 9 having larger rel-ative thicknesses. (iii) and (iv) Similarly, Log B lacks compositions 5, 6, and 10; and Log C lacks 2, 5, 7, 8, and 9 because of architecture.However, each log does record the overall increasing value of the x/y ratio, capturing broad trends in the system.

RATCLIFFE ET AL. 175

model is an oversimplification of the process ofdepositing and preserving alluvial strata, these gen-eral outcomes are realistic for any depositional allu-vial system.

When the larger spatial scale of the megafan andthe longer time scales involved in moving and depos-iting such large volumes of sediment are considered,it is clear that other factors will affect mineralogyrecorded in the sediments (Figure 13). These factorscan be grouped into three types: catchment factors,transport factors, and depositional basin factors. Inthe catchment bedrock lithology, climate, weathering,and rates of uplift and erosion all influence whichminerals are supplied to bedrock rivers, which con-vey sediment to the basin, and how much or how fastthis sediment is delivered (Figure 13). These arelinked to tectonic setting, latitude, and geography(e.g., proximity to coastlines and other mountainbelts). During transport, differential resistance to

abrasion by different minerals leads to downstreamchanges in mineral assemblages, which are amplifiedby selective sorting both in in-channel bed formsand during overbank splaying (off-axis sorting)(Figure 13). Sediments also occur in pulses, respond-ing to storm events, which take time to propagate intothe basin, and in the process mix with other sedimentsbuffering provenance signals. Once the sedimentshave come to rest in the basin, basin climate andweathering, subsidence rates, diagenesis, and meta-morphism, occurring during burial and exhumation,all could possibly alter mineralogy (Figure 13).These are just a few of the possible physical andchemical processes which could influence the miner-alogical makeup of a sedimentary deposit. All orsome of these processes could have acted on sedi-ments of the Abrahamskraal Formation studied here.

Another issue for geochemical correlation is theresultant scale of the megafan and overlapping of

Figure 13. Summary diagram illustrating the large-scale stacking arrangement envisaged for fluvial megafans by compensationalstacking. Two contemporaneous megafans are shown, the left one having three alluvial ridges depositing three lobes labelled A–C.Note, only river C is active. Hypothesized controls on chemostratigraphy are shown in uppercase. Topographic and depositional featuresare indicated in lowercase. Different megafans are shown by different shades of grey in strike and dip sections. Much like the stackingarrangement of alluvial ridges, the stacking of arrangements of fluvial megafans operates by compensational stacking. This means thatthe autogenic controls on geochemistry, related to avulsion, are very similar to the larger scale alluvial ridge (lobe) scale (i.e., prov-enance changes in the hinterland and selective sorting downstream and off-axis). However, the greater size of fluvial megafans to theircomponent alluvial rivers means that longer range changes in certain factors, such as weathering and climate, may have had a signifi-cant influence. Note that the distribution of channel belts within the megafan deposits are schematic only. Vertical scale is exaggerated.

176 Ground Truthing Chemostratigraphic Correlations in Fluvial Systems

adjacent megafans (Figure 13). It can be argued thatfor a fixed distance between samples, such as we havein our study, the larger the megafan, the smaller thevariability is likely to be between geochemical profilesbecause many components of the system are also big-ger and take longer to occur. However, insufficientstudies of modern rivers exist to clarify this issue, andit remains a theoretical problem. Overlapping of adja-cent megafans could introduce a large change in geo-chemistry, as seen in a geochemical log, because ofthe possibility of bedrock geology varying along thehinterland mountain belt, producing changing mineral-ogy of the sediment fed into the megafan along theorogeny (Figure 13). Clearly, this occurs, but thelength scale of variation will be variable.

Our description of these issues is necessarilyqualitative as they remain largely unknown, thoughcan be modelled numerically. Application of studiesfrom modern rivers to the ancient rock record is alsofraught with uncertainty about whether they can re-present rivers present in Gondwana (a continentgreater in size than anything today) and to such longperiods of time (∼5 m.y.). However, these factorshave been discussed to highlight the issues that aresurrounding interpretation of geochemistry of fluvialstrata and addressing the length scales over whichchemostratigraphic correlation can be considered via-ble in the rock record.

The thicknesses of fluvial stratigraphy in thisstudy (∼400 m [1312 ft]) are the product of ongoingavulsion over a period of approximately 4.7 m.y.,as determined by measured rates of deposition(85 m/m.y. [278.9 ft/m.y.]) by Lanci et al. (2013).If the thickness of an alluvial ridge is approximatelythe thickness of a single channel belt, averaging 5 m(16 ft) thick in this study, and Ouberg Pass containsabout 400 m (1312 ft) of fluvial strata, then in excessof 80 alluvial ridges could be represented at thislocation if vertically stacked, and probably more ascompensational stacking is a lateral process. Thus,the studied interval represents deposition from alarge number of rivers, probably originating fromthe same or adjacent catchments producing severalmegafans from the front of the mountain range(Figure 13) over 5 m.y. Simplistic modelling sug-gests that at the scale of a channel belt thickness(5 m [16 ft]), individual features of the

chemostratigraphic profiles may not appear on alladjacent logs and may differ in thickness in adjacentlogs. In contrast, at a thickness scale greater than thechannel belt (e.g., >25 m [> 82 ft]) chemostrati-graphic trends are likely similar, because at thisscale compositional stacking averages out these var-iations (Figures 12, 13). Thus, features found insome chemostratigraphy profiles are not expectedto be found in all adjacent profiles, and short andlong distance correlations are expected to showgreater dissimilarity.

CONCLUSIONS

For the example outlined here, correlation in fluvialsystems by chemostratigraphy is valid and is opti-mally achieved using fine-grained lithologies. Theshorter the distance between measured sections thegreater will be the chance of capturing a high numberof common features and producing a high-resolutioncorrelation. Coarse-grained lithologies (fluvial chan-nel belts) do show compositional packages, but theirsubordinate nature precludes their use for high-reso-lution correlation. Subdivision of strata through ahierarchical approach into sedimentary packagesbased on isochemical intervals allows chemostrati-graphic packages to be defined. Chemostratigraphiccorrelations have been shown to mirror physical cor-relations over distances up to 25.5 km (15.8 mi), butprovide the advantage of higher resolution subdivi-sions to be made based on objective data (geochemis-try) not captured by lithostratigraphic logging alone(e.g., grain size, sedimentary structures, etc.).

The elements and element ratios used to definethe chemostratigraphic correlation here reflectchanges in sediment provenance and weatheringstyle, both of which are to be expected in almostany fluvial system. Therefore, the limiting factor onthe technique of chemostratigraphy in a fluvialsystem is not likely to be a lack of chronostrati-graphically significant changes in elemental compo-sitions; instead, it will be a complex interplaybetween the rate at which changes occur relative tothe sedimentation rates, the architecture of the sys-tem, the distances over which correlation is beingattempted, and the scale of stratigraphic units that

RATCLIFFE ET AL. 177

are being sought. In the intervals discussed here, it isevident that sufficient changes in provenance andweathering style occurred to allow intervals tens tohundreds of meters thick to be correlated over dis-tances up to 25.5 km (15.8 mi). However, it isunlikely that a distance or resolution limit can bedefined universally for the effectiveness of chemo-stratigraphy as a correlation technique in fluvial sys-tems. Any such limitations will vary with the scaleof the fluvial system studied and the dynamics ofthe river system during deposition that would affectthe routing of sediment and the size of the fluvialmegafan. For most depositional systems, knowledgeof the size of the system is difficult to determinewith confidence.

Correlation and subdivision using chemostratig-raphy has a distinct advantage over lithostratigraphyalone as it is able to measure chemical properties thatcan then be used to identify unique geochemicalpackages of sediment in successions of otherwisenon-unique combinations of facies. The constructionof a facies scheme during the process of loggingrequires the pigeon-holing of what is realistically acontinuum of sedimentary features into specific lim-ited types that lose information about the strata.With the chemostratigraphy technique, individualelements are quantitatively measured, dispensingwith the need to summarize a bed of an outcrop intosemiquantitative or simplified measures such as aver-age grain size and local sedimentary structures.Chemostratigraphy has allowed detailed subdivisionof the logged sections into geochemical units, a levelof detail not achieved using physical correlation fromVOM and sedimentary logging alone.

The method used here for data acquisition has beenshown by numerous authors to be applicable to subsur-face samples, including cuttings samples (Hildredet al., 2010; Ratcliffe et al., 2010; Sano et al., 2013).The approaches and methods documented here usingfield samples can readily be exported to any fluvialsystem of any age, either at outcrop or in the subsur-face. The successful outcome in terms of providingchronostratigraphically significant correlations thatprovide value in an oil field setting cannot be quantifiedbut will largely depend on the scale of the fluvialsystem relative to the distances between sections andthe resolution of the correlation being sought.

REFERENCES CITED

Algeo, T. J., L. Schwark, and J. C. Hower, 2004, High-resolutiongeochemistry and sequence stratigraphy of theHushpuckney Shale (Swope Formation, eastern Kansas):implications for climato-environmental dynamics of theLate Pennsylvanian Midcontinent Seaway: ChemicalGeology, v. 206, p. 259–288, doi:10.1016/j.chemgeo.2003.12.028.

Armstrong, A. J. S., Y. I. Lee, S. P. Verma, and S. Ramasamy,2004, Geochemistry of sandstones from the upper MioceneKudankulam Formation, southern India: Implications forprovenance, weathering, and tectonic setting: Journal ofSedimentary Research, v. 74, p. 285–297, doi:10.1306/082803740285.

Bridge, J. S., and M. R. Leeder, 1979, A simulation model of allu-vial stratigraphy: Sedimentology, v. 26, p. 617–644, doi:10.1111/j.1365-3091.1979.tb00935.x.

Buckley, S. J., J. Vallet, A. Braathen, and W. Wheeler, 2008,Oblique helicopter-based laser scanning for digital terrainmodelling and visualisation of geological outcrops:International Archives of the Photogrammetry, RemoteSensing and Spatial Information Sciences, v. 37, Part B4,p. 493–498, doi:10.2110/jsr.2011.040.

Buehler, H. A., G. S. Weissmann, L. A. Scuderi, and A. J.Hartley, 2011, Spatial and temporal evolution of an avulsionon the Taquari River distributive fluvial system from satel-lite image analysis: Journal of Sedimentary Research,v. 81, no. 8, p. 630–640.

Davies, E. J., K. T. Ratcliffe, P. Montgomery, L. Pomar, B. B.Ellwood, and D. S. Wray, 2014, Magnetic susceptibilitystratigraphy and chemostratigraphy applied to an isolatedcarbonate platform reef complex, Llucmajor, in K. Verwer,T. Playton, and P. Harris, eds., Architecture andControls of Carbonate Margin, Slope, and Basin Systems:SEPM Special Research Symposium, SEPM SpecialPublication.

DeCelles, P. G., and W. Cavazza, 1999, A comparison of fluvialmegafans in the Cordilleran (Upper Cretaceous) and modernHimalayan foreland basin systems: Geological Society ofAmerica, Bulletin, v. 111, no. 9, p. 1315–1334, doi:10.1130/0016-7606(1999)111<1315:ACOFMI>2.3.CO;2.

Demko, T. M., B. S. Currie, and K. Nicoll, 2004, Regionalpaleoclimatic and stratigraphic implications of paleosolsand fluvial/overbank architecture in the MorrisonFormation (Upper Jurassic), Western Interior, USA:Sedimentary Geology, v. 167, p. 115–136, doi:10.1016/j.sedgeo.2004.01.003.

Ellis, D. V., and M. Singer, 2007, Well logging for EarthScientists: Dordrecht, The Netherlands, Springer Science +Business Media B.V., 692 p.

Ellwood, B. B., J. H. Tomkin, K. T. Ratcliffe, A. M. Wright, andA. M. Kafafy, 2008, Magnetic susceptibility and geochemis-try for the Cenomanian/Turonian boundary GSSP withcorrelation to time equivalent core: Palaeogeography,Palaeoclimatology, Palaeoecology, v. 251, p. 1–22.

English, L. T. P., 1999, The use of magnetic susceptibility andtrace element geochemistry for the correlation of finegrained siliciclastic sequences: a Late Llandovery example

178 Ground Truthing Chemostratigraphic Correlations in Fluvial Systems

from northwest England: Geological Magazine, v. 136,no. 04, p. 423–436, doi:10.1017/S0016756899002587.

Flint, S., D. Hodgson, and A. Sprague, 2011, Depositional archi-tecture and sequence stratigraphy of the Karoo basin floor toshelf edge succession, Laingsburg depocentre, South Africa:Marine and Petroleum Geology, v. 28, no. 3, p. 658–674,doi:10.1016/j.marpetgeo.2010.06.008.

Fontana, A., G. Monegao, E. Zavagno, S. Devoto, I. Burla, andF. Cucchi, 2013, Evolution of an Alpine fluvioglacial sys-tem at the LGM decay: The Cormor Megafan (NE Italy):Geomorphology, v. 204, p. 136–153.

Gupta, S., 1997, Himalayan drainage patterns and the origin offluvial megafans in the Ganges foreland basin: Geology,v. 25, no. 1, p. 11–14, doi:10.1130/0091-7613(1997)025<0011:HDPATO>2.3.CO;2.

Haughton, P. D. W., S. P. Todd, and A. C. Morton, 1991,Sedimentary provenance studies: Geological Society ofLondon, Special Publications, v. 57, p. 1–11, doi:10.1144/GSL.SP.1991.057.01.01.

Hildred, G. V., K. T. Ratcliffe, A. M. Wright, B. A. Zaitlin, andD. S. Wray, 2010, Chemostratigraphic applications to low-accommodation fluvial incised-valley settings; an examplefrom the Lower Mannville Formation of Alberta, Canada:Journal of Sedimentary Research, v. 80, p. 1032–1045,doi:10.2110/jsr.2010.089.

Horton, B. K., and P. G. DeCelles, 2001, Modern and ancient flu-vial megafans in the foreland basin system of the centralAndes, southern Bolivia: implications for drainage networkevolution in fold-thrust belts: Basin Research, v. 13, no. 1,p. 43–63, doi:10.1046/j.1365-2117.2001.00137.x.

Jarvis, I., and K. E. Jarvis, 1995, Plasma spectrometry in earth sci-ences: techniques, applications and future trends, in Jarvis,I., Plasma Spectrometry in Earth Sciences: ChemicalGeology, v. 95, p. 1–33, doi:10.1016/0009-2541(92)90041-3.

Johnson, M. R., C. J. Van Vuuren, J. N. J. Visser, D. I. Cole,H. de V. Wickens, A. D. M. Christie, and D. L. Roberts,1997, The foreland Karoo Basin, South Africa, in K. J.Hsü, ed., African basins—sedimentary basins of the world,Amsterdam, Elsevier, p. 269–317.

Lanci, L., E. Tohver, A. Wilson, and S. Flint, 2013, UpperPermian magnetic stratigraphy of the lower BeaufortGroup, Karoo Basin: Earth and Planetary Science Letters,v. 375, p. 123–134, doi:10.1016/j.epsl.2013.05.017.

Mohindra, R., B. Parkash, and J. Prasad, 1992, Historicalgeomorphology and pedology of the Gandak megafan,Middle Gangetic Plains, India: Earth Surface Processesand Landforms, v. 17, no. 7, p. 643–662, doi.:10.1002/(ISSN)1096-9837.

Morton, A., C. Hallsworth, J. Kunka, E. Laws, S. Payne, and D.Walder, 2010, Heavy mineral stratigraphy of the ClairGroup (Devonian—Carboniferous) in the Clair Field, Westof Shetland, UK, in K. T. Ratcliffe, ed., Modern alternativestratigraphic techniques; theory and case histories, SEPMSpecial Publication No. 94, p. 183–201.

Nichols, G. J., and J. A. Fisher, 2007, Processes, facies and archi-tecture of fluvial distributary system [936] deposits:Sedimentary Geology, v. 195, no. 1–2, p. 75–90, doi:10.1016/j.sedgeo.2006.07.004.

North, C. P., M. J. Hole, and D. G. Jones, 2005, Geochemical cor-relation in deltaic successions; a reality check: GeologicalSociety of America Bulletin, v. 117, no. 5–6, p. 620–632,doi:10.1130/B25436.1.

North, C. P., G. C. Nanson, and S. D. Fagan, 2007, Recognitionof the sedimentary architecture of dryland anabranching(anastomosing) rivers: Journal of Sedimentary Research,v. 77, no. 11, p. 925–938, doi:10.2110/jsr.2007.089.

Pe-Piper, G., S. Triantafyllidis, and D. J. W. Piper, 2008,Geochemical identification of clastic sediment provenancefrom known sources of similar geology: the CretaceousScotian Basin, Canada: Journal of Sedimentary Research,v. 78, p. 595–607, doi:10.2110/jsr.2008.067.

Pearce, T. J., D. S. Wray, K. T. Ratcliffe, D. K. Wright, and A.Moscariello, 2005a, Chemostratigraphy of the UpperCarboniferous Schooner Formation, southern North Sea, inJ. D. Collinson, D. J. Evans, D. W. Holliday, and M. S.Jones, eds., Carboniferous Hydrocarbon Geology: TheSouthern North Sea and Surrounding Onshore Areas:Yorkshire Geological Society, Occasional Publicationsseries, v. 7, p. 147–164.

Pearce, T. J., D. McLean, D. S. Wray, D. K. Wright, C. J. Jeans,and E. W. Mearns, 2005b, Stratigraphy of the UpperCarboniferous Schooner Formation, southern North Sea: che-mostratigraphy, mineralogy, palynology and Sm-Nd isotopeanalysis, in J. D. Collinson, D. J. Evans, D. W. Holliday, andM. S. Jones, eds., Carboniferous Hydrocarbon Geology: TheSouthern North Sea and Surrounding Onshore Areas:Yorkshire Geological Society, Occasional Publications series,v. 7, p. 165–182.

Pearce, T. J., J. H. Martin, D. Cooper, and D. S. Wray, 2010,Chemostratigraphy of Upper Carboniferous (Pennsylvanian)sequences from the southern North Sea (United Kingdom),in K. T. Ratcliffe, ed., Modern alternative stratigraphic tech-niques; theory and case histories: SEPM Special PublicationNo. 94, p. 109–129.

Ratcliffe, K. T., A. D. Hughes, D. E. Lawton, D. S. Wray,F. Bessa, T. J. Pearce, and J. Martin, 2006, A regionalchemostratigraphically-defined correlation frameworkfor the late Triassic TAG-I in Blocks 402 and 405a,Algeria: Petroleum Geoscience, v. 12, p. 3–12, doi:10.1144/1354-079305-669.

Ratcliffe, K. T., A. M. Wright, P. Montgomery, A. Palfrey, A.Vonk, J. Vermeulen, and M. Barrett, 2010, Application ofchemostratigraphy to the Mungaroo Formation, the GorgonField, offshore Northwest Australia: APPEA Journal 201050th Anniversary Issue, p. 371–388.

Ratcliffe, K. T., A. M. Wright, and K. Schmidt, 2012a,Application of inorganic whole rock geochemistry to shaleresource plays: an example from the Eagle Ford ShaleFormation, Texas: Sedimentary Record, v. 10, p. 4–9.

Ratcliffe, K. T., A. M. Wright, and D. Spain, 2012b,Unconventional methods for unconventional plays: usingelemental data to understand shale resource plays. PESANews Resources, March/April 2012, p. 89–93.

Retallack, G. J. 1997, A colour guide to palaeosols: Chichester,John Wiley and Sons, 175 p.

Rittersbacher, A., S. J. Buckley, J. A. Howell, G. J. Hampson, andJ. Vallet, in press, Helicopter-based laser scanning: a method

RATCLIFFE ET AL. 179

for quantitative analysis of large-scale sedimentary architec-ture, in, A. W. Martinius, J. A. Howell, and T. Good, eds.,Sediment body geometry and heterogeneity: London,Geological Society, Special Publication, 387 p.

Sano, J. L., K. T. Ratcliffe, and D. R. Spain, 2013,Chemostratigraphy of the Haynesville Shale, in U. Hammes,Geology of the Haynesville Gas Shale in East TexasandWest Louisiana, U.S.A.: AAPGMemoir 105, p. 137–154.

Schoch, R. N. M., 1989, Stratigraphy: Principles and Methods:London, Chapman and Hall, 375 p.

Singh, H., B. Parkash, and K. Gohain, 1993, Facies analysis ofthe Kosi megafan deposits: Sedimentary Geology, v. 85,no. 1, p. 87–113, doi:10.1016/0037-0738(93)90077-I.

Smith, R. M. H., 1990, Alluvial paleosols and pedofacies sequen-ces in the Permian Lower Beaufort of [967] the southwesternKaroo Basin, South Africa: Journal of SedimentaryResearch, v. 60, no. 2, p. 258–276.

Smith, R. M. H., 1998, Sedimentology and ichnology of acrevasse splay palaeosurface in the Beaufort Group (LatePermian), Karoo Supergroup, South Africa: Journal ofAfrican Earth Sciences, 27, no. 1, p. 184–184.

Stanistreet, I. G., and T. S. McCarthy, 1993, The Okavango Fanand the classification of subaerial fan systems: SedimentaryGeology, v. 85, no. 1, p. 115–133, doi:10.1016/0037-0738(93)90078-J.

Stear, W. M., 1978, Sedimentary structures related to fluctuatinghydrodynamic conditions in flood plain deposits ofthe Beaufort Group near Beaufort West, Cape:Transactions of the Geological Society of South Africa,v. 81, p. 393–399.

Svendsen, J., H. Friis, H. Stollhofen, and N. Hartley, 2007,Facies discrimination in a mixed fluvio-eolian settingusing elemental whole-rock geochemistry: applicationsfor reservoir characterization: Journal of Sedimen-tary Research, v. 77, p. 23–33, doi:10.2110/jsr.2007.008.

Thomas, W. A., 2011, Detrital-zircon geochronology and sedi-mentary provenance: Lithosphere, v. 3, p. 304–308, doi:10.1130/RF.L001.1.

Torres, M. A., and R. R. Gaines, 2013, Paleoenvironmental andpaleoclimatic interpretations of the Late Paleocene GolerFormation, Southern California, U.S.A., based on paleosolgeochemistry: Journal of Sedimentary Research, v. 83,p. 591–605, doi:10.2110/jsr.2013.48.

Van de Velde, J. H., G. J. Bowen, B. H. Passey, and B. B. Bowen,2013, Climatic and diagenetic signals in the stable isotopegeochemistry of dolomitic paleosols spanning thePaleocene–Eocene boundary: Geochemica et CosmochimicaActa, v. 109, p. 254–267, doi:10.1016/j.gca.2013.02.005.

Weissmann, G. S., A. J. Hartley, G. J. Nichols, L. A. Scuderi,M. Olson, H. Buehler, and R. Banteah, 2010, Fluvialform in modern continental sedimentary basins: Distributivefluvial systems: Geology, v. 38, no. 1, p. 39–42, doi:10.1130/G30242.1.

Wickens, H. de V., 1994, Basin Floor Fan building Turbidites ofthe South-western Karoo Basin, Permian Ecca Group, SouthAfrica: Unpublished Ph.D. thesis, University of PortElizabeth, 233 p.

Wilson, A., S. Flint, T. Payenberg, E. Tohver, and L. Lanci, 2014,Architectural styles and sedimentology of the fluvial lowerBeaufort Group, Karoo Basin, South Africa: Journal ofSedimentary Research, v. 84, p. 326–348, doi:10.2110/jsr.2014.28.

Wright, A. M., K. T. Ratcliffe, B. A. Zaitlin, and D. S. Wray,2010, The application of chemostratigraphic techniques todistinguish compound incised valleys in low-accommodationincised-valley systems in a foreland-basin setting: an examplefrom the Lower Cretaceous Mannville Group and BasalColorado Sandstone (Colorado Group), Western CanadianSedimentary Basin, in Ratcliffe, K. T., ed., Modern alterna-tive stratigraphic techniques; theory and case histories,SEPM Special Publication No. 94, p. 93–109.

180 Ground Truthing Chemostratigraphic Correlations in Fluvial Systems