Available online at www.sciencedirect.com

Organic Geochemistry 39 (2008) 1150–1161

www.elsevier.com/locate/orggeochem

OrganicGeochemistry

Oil–source rock correlations – Limitationsand recommendations

J.A. Curiale *

Chevron Energy Technology Company, 14141 Southwest Freeway, Sugar Land, TX 77479, United States

Received 21 August 2007; received in revised form 8 January 2008; accepted 3 February 2008Available online 9 February 2008

Abstract

A rigorous, defensible and causal oil–source rock correlation definitively ties an individual source rock sample to anindividual crude oil using genetically-based, internally consistent parameter matches. Such a correlation is a relationshipestablished between the two components which is consistent with all known chemical, geochemical and geological infor-mation, and meets three criteria: (a) the relationship must be causal – the oil must arise (at least in part) from the specifiedsource rock(s); (b) chemical data used in the correlation must be comparable – the elemental, molecular and isotopic dataderived from the source rock must be of the same type as that derived from the oil; and (c) all available geological datamust be supportive – clear geological evidence must exist which allows the proposed source rock to have sourced the oil.The three points of this definition are satisfied, and the correlation is successful, if we solve one analytical and three geo-logical problems. Natural extraction of crude oil from its source rock – i.e., ‘‘expulsion” – differs mechanically and tem-porally from laboratory extraction, leading to a correlation problem referred to as ‘‘extraction differences”, andrepresenting the single greatest analytical uncertainty in correlation efforts. Geologically, three aspects confound oil–sourcecorrelations: occurrence of multiple source units and/or source units at differing maturity levels; lateral and vertical depo-sitional variations in source unit(s) organic matter; and lateral and vertical variation in-reservoir unit(s) organic matter.

These concerns are overcome by taking five actions for each oil–source correlation effort. (1) Select representative sam-ples using statistically defensible methods. (2) Establish the inherent compositional variability – laterally and temporally –due to depositional and maturation processes in each prospective source unit. (3) Assess the extent of migration-inducedchanges in oil composition, including post-migration changes such as in-reservoir alteration. (4) Support each correlationwith migration histories derived from 4d models. (5) Iterate correlation results with new data gathered from ongoing explo-ration efforts. These five actions will (a) allow correlation success to be measured more objectively, (b) establish riskingparameters for direct use in basin assessment, and (c) provide baseline criteria for use in assessing the reliability of oil–source rock correlations.

Although these approaches will improve conventional oil–source rock correlation efforts, ultimately they may not beneeded if inversion methods become more reliable. The ability to invert compositional data for crude oil – i.e., predictthe age, lithology, maturity, and other characteristics of a source rock solely through chemical analysis of its expelledoil – could eventually result in successful oil–source rock ‘‘correlations” in which source rock analysis for oil–source rockcorrelation purposes is less critical.� 2008 Elsevier Ltd. All rights reserved.

0146-6380/$ - see front matter � 2008 Elsevier Ltd. All rights reserved.doi:10.1016/j.orggeochem.2008.02.001

* Tel.: +1 281 287 5646; fax: +1 281 276 9351.E-mail address: jcuriale@chevron.com

J.A. Curiale / Organic Geochemistry 39 (2008) 1150–1161 1151

1. Introduction

Correlation of a crude oil to one or more sourcerocks is a common industrial application of petro-leum geochemistry. Confirmation that oil has beengenerated in the target sedimentary basin is the mostcritical piece of knowledge a petroleum explora-tionist can derive; second in importance is the deter-mination of the source(s) of that oil. For thisreason, an extensive arsenal of analytical methodsis utilized to collect primary data on the organicmatter in crude oils and possible source rocks, andvarious components of these data are used to relateoils causally to their prospective sources. Oil–sourcerock correlations at various confidence levels havebeen established for the petroleum systems of allmajor sedimentary basins.

Published case studies establishing causal corre-lations between oil and source rock pairs began toappear in the third quarter of the 20th century(e.g., Hunt et al., 1954; Welte, 1965; Dow, 1974;Williams, 1974) and summaries of oil–source rockcorrelation studies and methods are now available(Curiale, 1993, 1994a; Waples and Curiale, 1999).Although correlations are fundamentally subjectiveexercises, the results are rarely challenged eitherconceptually or on their merits. Numerous pub-lished examples correlating one or two oil samplesto one or two source rocks purport to establish rela-tionships which are commonly carried basinwide.Other examples utilize a decidedly limited suite ofgeochemical analytical types and data, often estab-lishing strong chemical correlations but no convinc-ing causality. Still others attempt to enforce achemical correlation on sample sets whose geologiccircumstances make such a correlation quiteunlikely.

Concerns of this type have been expressed sinceoil–source correlations were first attempted,although often only as passing remarks in publishedliterature and industry reports. Interestingly,although these concerns have been duly noted byvarious authors, most of these authors proceededwith their correlation efforts anyway, without anyserious effort to address the noted concerns. Thishas led to unnecessary uncertainty in publishedresults, and the impression that any correlation isbetter than no correlation. My objectives here areto gather into a single publication the reasons forthe subjectivity of oil–source correlations and torecommend conceptual and practical improvementsfor the use of these correlations in solving industrial

problems. In addition, I present some criteria forjudging the success of practical oil–source rock cor-relations, in the hope that these criteria will assist indefining the risk associated with the charge compo-nent of exploration efforts in both frontier andmature basins.

1.1. Requirements of an oil–source rock correlation

Any successful oil–source rock correlation mustinclude three attributes: (a) requirement of causal-ity; (b) comparable chemical data for all samples;and (c) geological support. For the purposes of thispaper, I will use the definition of Curiale (1993),derived from previous authors and cited in that pub-lication as: an oil–source rock correlation is a causal

relationship, established between a crude oil and an

oil-prone petroleum source rock, which is consistent

with all known chemical, geochemical and geological

information. The three key points of this definitionare listed below.

� The relationship must be causal. That is, the oilmust arise (at least in part) from the specifiedsource rock.� Chemical data used in the correlation must be

comparable. That is, the elemental, molecularand isotopic data derived from the source rockmust be of the same type as that derived fromthe oil.� All available geological data must be supportive.

That is, clear geological evidence must existwhich allows the proposed source rock to havesourced the oil.

The absence of any of these three key points nec-essarily negates the validity of a proposed oil–sourcerock correlation. That is: the presence of all threepoints is required, at a minimum, before declaringa correlation to be successful.

The importance of both the chemical and geolog-ical character of these three definitional points can-not be overemphasized. Establishing chemicalsimilarities between the organic matter in a sourcerock and that in an oil, even if these similaritiesinvolve ‘genetic’ (i.e., source-derived) molecularand isotopic characteristics, is necessary but insuffi-cient. Such a result must also be supplemented bysupporting geological data establishing that thesource was capable – in all spatial and temporaldimensions – of having generated a specific oil.These geological data, including the details of depo-

1152 J.A. Curiale / Organic Geochemistry 39 (2008) 1150–1161

sitional history and structural configurationthrough time, are provided as input to a robustbasin model which is used to support the correlationconclusion in the spatial (i.e., fluid flow configura-tion) and temporal (i.e., timing of generation andexpulsion) dimensions. Only when this is confirmedcan a bona fide oil–source rock correlation be con-cluded with confidence.

1.2. Previous approaches

Several published, successful oil–source rock cor-relations have appeared since the first was presentedby John Hunt and colleagues over 50 years ago.Hunt et al. (1954) provided clear chemical and geo-logical evidence in a causal framework for thesource of oils and solid bitumens of the Uinta Basin(USA). Two decades passed before publication ofthe seminal work by Williams (1974) and Dow(1974) on the Williston Basin (USA), in which ele-mental, isotopic and molecular data were used insuccessful correlations among three source rocksand oil families. Although later studies became pro-gressively more analytically sophisticated, the con-cepts established by Hunt, Williams and Dow,from which the definition given above was derived,still dominate oil–source rock correlationapproaches.

All analytical methods used by petroleum geo-chemists to characterize oils and the organic matterin source rocks have also been used to correlate oilsto source rocks, and the confidence we have in oil–source correlations parallels the confidence we havein petroleum geochemical methods in general.Indeed, the ease and rapidity with which precise ana-lytical data can now be obtained has tended to de-emphasize the importance of geological data,because the precision of chemical data tends to befar greater than that of geological data. In part, thisis because rapid temporal changes in the deposi-tional conditions of many sedimentary systems makeit difficult to sample these systems representatively.Thus, the fundamental non-analytical obstacle tosuccessful oil–source rock correlations is the naturalvariability of the geological system, and much of thispaper focuses on coping with this variability whenattempting oil–source rock correlations.

1.3. The problem

Our knowledge of variability within depositionalsystems at all scales has not kept pace with our abil-

ity to generate highly precise analytical data.Indeed, chemical correlations are now easy to estab-lish – almost automatic when using some statisticalapproaches – whereas knowledge of natural, fine-scale variation in source rock deposition (e.g., Curi-ale, 1994b; Keller and Macquaker, 2001; Barkeret al., 2001) is substantially more difficult to acquire.Therefore, although a few aspects of analyticaluncertainty remain (mostly based on the type ofrock extraction method used – see below), the suc-cess of an oil–source rock correlation depends lar-gely on our understanding of sedimentary organicvariability in the natural setting.

This problem – making certain that the naturalvariability of depositional processes is fully consid-ered when establishing oil–source rock correlations– requires that lateral and vertical differences inorganic matter distribution in both source and res-ervoir rocks are accounted for properly. This vari-ability extends from organic differences in sourceshales at the laminae scale to the potential for multi-ple source units to drain toward a single focal point(trap). The remainder of this paper discusses thesefactors using examples. Individual sections brieflyaddress correlation problems arising from theoccurrence of extraction technique differences, mul-tiple source units, lateral/vertical source organicvariability and lateral/vertical reservoir organic var-iability. A closing section proposes recommenda-tions for more accurate and consistentcorrelations, in an effort to minimize the uncertaintyin oil–source correlations. In this manner, moreaccurate correlations can be used in an industrialcontext, thus reducing exploration risk.

2. Extraction differences

Most correlation techniques in common useinvolve molecular and isotopic comparison of com-ponents in crude oil with components extractedfrom a candidate source rock. Natural extractionof crude oil from its source rock – usually denotedas ‘‘expulsion” – differs chemically, mechanicallyand temporally from laboratory extraction, leadingto a correlation problem referred to here as ‘‘extrac-tion difference”. Laboratory extraction usually uti-lizes an organic solvent at its boiling point, occursover a convenient (i.e., short) time period in a man-ner that is relatively well controlled and understood(Eglinton and Murphy, 1969), and usually con-cludes with a time-efficient method of removingmost of the unwanted solvent. In contrast, natural

J.A. Curiale / Organic Geochemistry 39 (2008) 1150–1161 1153

extraction (expulsion) of oil from a source rockoccurs at greatly different timescales, via mecha-nisms that are much less well understood. Neverthe-less, the standard approach of the petroleumgeochemist is to relate molecular and isotopic char-acteristics of an oil with those in organic matter thathas been solvent-extracted from a source rock, whilemaking an effort (either implicit or explicit) not toutilize components thought to be differentiallyaffected by these two modes of extraction (e.g., thelightest components which, although often the bulkof the soluble organic matter in a source rock, areusually lost during sampling or laboratory workupextraction procedures). This is clearly a fundamen-tal analytical constraint in traditional oil–sourcerock correlation approaches. Although some work-ers have attempted to circumvent this concern byusing extraction fluids thought to be more relevantto those present in the subsurface and taking specialcare to retain native components during workup(e.g., through extraction with supercritical carbondioxide; Li et al., 1996; Koel et al., 2000), these lab-oratory approaches still suffer from a lack of simi-larity to natural expulsion. Indeed, mostcorrelations continue to use standard solvent extrac-tion techniques, leading to an unwarranted correla-tional focus on tetra- and pentacyclic hydrocarbonsbecause of their favorable workup retention charac-teristics. As a result, the approach represents thesingle greatest analytical uncertainty in correlationefforts, and will likely remain so.

Because the extent and nature of fractionationdifferences that occur between solvent extraction inthe laboratory and natural extraction during oilexpulsion are well known and extensive at the com-pound class level (e.g., Leythaeuser et al., 1984), afew workers have sought alternatives. A straightfor-ward approach to more closely mimicking naturalgeneration and expulsion is to apply pressure and/or temperature to prospective source intervals inthe laboratory, thereby artificially generating anoil for comparison with natural crude oil. Earlyefforts in this regard involved mechanically com-pacting organic-rich recent sediments, which led toexpulsion of hydrocarbon-rich fluids subsequentlyanalyzed and shown to be oil-like (e.g., Shinnet al., 1984). Because most petroleum source unitsare already lithified, these mechanical compactionefforts involving recent sediment have been replacedby various types of confined and unconfined heatingin an oxygen-free atmosphere. Of these pyrolyticapproaches, confined heating in the presence of

excess water commonly yields the most oil-likeexpelled fluid (Lewan, 1985). Nevertheless, detailedmolecular differences between natural oil and pyro-lyzate are still observed, limiting the approach tosome extent. Furthermore, the time- and labor-intensive nature of the hydrous pyrolysis approachlimits its use to non-routine applications, leavingus with continued concerns about extraction differ-ences when attempting oil–source rock correlations.

3. Mixing from multiple source units and multiple

maturity levels

Increasing use among petroleum geologists andgeochemists of thermal and flow modeling at thebasin scale has resulted in an increased awarenessof the occurrence and importance of multiple sourcerock units in a sedimentary basin, and the post-source mixing of oil expelled from these units enroute to the trap. In addition, maturity-inducedcompositional variations in crude oil and in the sol-uble organic matter remaining in a source rock afterexplusion can be significant, and efforts to correlatea crude oil with the residual soluble organic matterin its source rock(s) represent an inherent and fun-damental uncertainty in oil–source rock correlation.It is now recognized that few commercial oil and gasaccumulations are singly sourced, and even fewerare sourced from a single unit at a single maturitylevel. Examples of these sorts of mixing range fromthe classic work of Seifert et al. (1979) on the oils ofthe Prudhoe Bay Field (Alaska) to a large body ofwork published over the past two decades for sedi-mentary basins worldwide (e.g., Peters et al., 1989;Dzou et al., 1999; Chen et al., 2003a,b and refer-ences therein). Although the correlation problemsassociated with maturity differences can sometimesbe alleviated through judicious choice of rock sam-pling, oils mixed from sources of differing ages –usually at different maturity as well – confoundthe problem significantly.

Chen et al. (2003a,b), in their extensive study ofthe triply-sourced oils of the Cainan Field of Chi-na’s Junggar Basin, have established multiple sourceunit input by using molecular and isotopic mixingevidence as well as by creating artificial mixturesof inputs from Jurassic, Triassic and Permian sourcerocks. Potential variations in contributions fromthese three sources (Chen et al., 2003b) could resultin carbon isotope ratio differences of greater than4‰, pristane/phytane ratios varying by a factor oftwo, and relative concentrations of b-carotane vary-

1154 J.A. Curiale / Organic Geochemistry 39 (2008) 1150–1161

ing by more than an order of magnitude (Fig. 1).Chen et al. (2003a,b) clearly demonstrate that allthree sources are required as contributors to theoil in the Cainan Field, and show the wide rangeof single-field oil compositions which can be antici-pated when multiple sources are involved.

The contribution of expelled products from twoor more source units within a single sedimentaryfetch area presents a fundamental problem for anyoil–source correlation attempt. Accurate correlationof oil contained within a single trap (it is usuallyassumed that the oil is uniform throughout thetrapped volume) and derived from multiple sourcesrequires detailed compositional information foreach of the sources, as well as information aboutlateral and vertical compositional variability in eachsource due to depositional (the importance of thisitem is discussed below) and maturity differences.Furthermore, the extent of mixing (as well as thetime and place of mixing) will affect the compositionof the resultant oil. Because most of these aspectsare either unknown or poorly known, oil–sourcecorrelations involving multiple source units and/ormaturity levels possess increased risk. This riskcan be alleviated to some extent by examining thefull molecular weight spectrum of an oil (ratherthan, say, just the tetracyclic and pentacyclic bio-

Fig. 1. Variation in b-carotane/n-C25 hydrocarbon ratio for oils of themixtures of Permian-sourced, Triassic-sourced and Jurassic-sourced oilsof the field and of the mixtures. The hatchured rectangle shows the rahorizontal line within this rectangle shows the average of the ratio forshow the variability in b-carotane/n-C25 ratio as Permian oil (top curve)this are useful in estimating source inputs to multiply sourced oils. Th

marker components). The recent use of multiplemolecular fractions has yielded more robust correla-tions in several instances (Rooney et al., 1998;Waples and Curiale, 1999; Obermayer et al., 2000).

Oil–source correlation concerns arising frommultiple source inputs are exacerbated even furtherwhen widely varying component concentrations areused to support multiple sourcing, as exemplified bythe question of source components for the oils ofBeatrice Field, offshore Scotland. The Jurassic-res-ervoired oils of this field contain b-carotanes, whichare used by both Peters et al. (1989) and Bailey et al.(1990) to support a correlation to a Devonian lacus-trine source unit. In addition, however, the presenceof low concentrations of n-propyl cholestanes areused by Peters et al. (1989) to infer a secondary,Jurassic source input. This co-sourcing conclusionis refuted by Bailey et al. (1990), who state that‘‘as a matter of philosophy, we would suggest thatit is dangerous to base hypotheses upon componentswhich occur in petroleum only in trace amounts”

(cf. Curiale, 1994a). Although Peters et al. (1999)cite additional, supporting information for a Juras-sic co-source, the multiple-sourcing aspect of oil–source correlation efforts will continually raise thequestion: What molecular concentrations shouldwe require in an oil before we must seek those com-

Cainan Field, Junggar Basin (NW China; see inset) and artificial. The x-axis shows the percentage of Jurassic-sourced oil in the oilsnge in b-carotane/n-C25 ratio for the Cainan Field oils, and theall Cainan oils. The curves sloping from top left to bottom rightor Triassic oil (bottom curve) is added to Jurassic oil. Displays likeis figure is modified after figures in Chen et al. (2003a,b).

J.A. Curiale / Organic Geochemistry 39 (2008) 1150–1161 1155

ponents in a source? For example, are biomarker-based correlations reliable when a biomarker-richfluid (e.g., a low-maturity oil) mixes with a bio-marker-poor fluid (e.g., a thermal condensate)? Ashighly selective and sensitive mass spectrometricdetection methods have become more common-place, the answer to this question becomes morecritical, and represents a key concern when attempt-ing oil–source rock correlations in basins containingmultiple source units or in source units at multiplematurity levels.

4. Lateral and vertical organic variability in source

deposition

Inspection of modern depositional settings indi-cates that organic matter accumulation in fine-grained sediments can show extensive composi-

aliphatic gas chromatogram

n-C27 n-C29

n-C25

n-C31

Fig. 2. Molecular distributions in two sidewall core rock extracts fromlithology, and separated by less than 100 m in a lithologically-constant sis shown on the left (shallower sample at the top); prominent odd-carbalkanes eluting in between (not annotated). Note the differences in thealkane dominance in the deeper sample. Note also the obvious differenceof the m/z 410 mass chromatogram (GC-MSD data) for each samplecomponents, including oleanenes with unsaturation at the 12, 18 andoleanenes in the two chromatograms, despite constant lithology and maminor (angiospermous?) components in the deeper (lower) sample.

tional variability, both laterally and temporally(Tyson, 1995). As modern sediments become partof the lithified stratigraphic section, this variabilityis often preserved as compositional variations atthe bulk (organic petrographic) and molecular lev-els. Lithologically homogenous sections often showmolecular differences over relatively small verticalextents. An example of this is shown in Fig. 2.Two Oligocene rock extracts from northernAlaska, sampled as sidewall cores and separatelyvertically by less than 100 m (and thus at essentiallyidentical thermal maturity level), are very similar inbulk composition – TOC = 1.16% and 1.26%;d13Caliph = �28.4‰ and �28.6‰. Yet, these sam-ples show measurable and significant differencesin acyclic and cyclic biomarker distributions (seechromatograms in Fig. 2, as well as additionaldetails in the figure caption).

m/z 410 mass chromatogram

Δ12, Δ18, Δ13(18)

-oleanenes

northern Alaska. Both samples are early Oligocene, of identicaltratigraphic section. A portion of the aliphatic gas chromatogramon number n-alkanes are annotated, with even-carbon number n-distribution of the n-alkanes, including the greater odd-carbon n-s in non-n-alkane distribution between the two samples. A portion

is shown at the right; a cluster of olefinic angiosperm-derived13(18) positions, is annotated. Note the differing distribution ofturity. Note also the apparent occurrence of numerous additional,

1156 J.A. Curiale / Organic Geochemistry 39 (2008) 1150–1161

Although the example in Fig. 2 shows variabilityat the 100 m scale, studies at much higher samplingresolution also consistently reveal vertical organicvariability, often at the centimeter scale (e.g.,Fig. 3), that is as great as that observed at the meterand kilometer scales (Huc et al., 1992; Aplin et al.,1992; Curiale, 1994b; Burwood et al., 1995; Joneset al., 1997; Nijenhuis et al., 1999; Keller and Mac-quaker, 2001; Barker et al., 2001). Furthermore, theextent of this variability can change significantlyfrom one depositional setting to another. For exam-ple, deltaic source rocks containing Type II/III orIII kerogen show much greater variability, both lat-erally and vertically, than freshwater lacustrinesource rocks containing Type I kerogen (cf. Giblinget al., 1985; Lin et al., 2005). An example of this isthe Serravallian (mid-Miocene) section of coastalCalifornia, where 1 m and 10 m intervals of theMonterey Formation show elemental, molecularand isotopic variability equivalent to that of theentire vertical succession of the Monterey Forma-tion (Curiale, unpublished data).

As inferred earlier, the composition of crude oilmay be envisioned as a time-averaged productexpelled from all candidate sources within a single

Fig. 3. Total organic carbon variability in a 40 cm sectioncovering approximately 40,000 yr of deposition in the EarlyAlbian of southeastern United States (modified after Barkeret al., 2001). Note the large (more than an order of magnitude)and rapid variation in organic carbon preservation. Largevariations are also observed in hydrogen index (99–790 mg/g)and oxygen index (33–394 mg/g) in this 40 cm section.

fetch area. Just as multiple source units can con-found oil–source rock correlations, temporally andlaterally rapid internal differences in the organicfacies of each source unit (that is, the lateral andvertical changes in kerogen type within a singlesource unit) will do the same. The temporal scaleof these changes can be decadal or even annual; lat-eral changes can occur over as little as meters or tensof meters. As a result, the integrated product ofexpulsion from these various facies – i.e., the oilexpelled from the source unit – is likely to be com-positionally unlike that expelled from any individ-ual sample from within the source sequence. Anexample of this is shown in Fig. 4, where the isoto-pic distribution of n-alkanes in a Camamu Basin(Brazil) oil is unlike that of any sample of its pur-ported source unit, but nearly identical to that ofa weighted average of four prospective samplesfrom the source unit (Curiale and Sperry, 2000).As finer and finer analytical methods are utilized(in this case, compound specific stable isotope anal-ysis for carbon), the observation of mismatchesbetween individual rock samples and crude oils arelikely to become more commonplace.

5. Organic variability in carrier bed and reservoir

rock

Although organic facies variations in oil-pronesource units provide a first-order control on thecomposition of expelled oil, so-called non-geneticor post-sourcing events can also change an oil’scomposition. In-trap biodegradation, for example,results in several well-recognized compositionalchanges in crude oil (Larter et al., 2006, and refer-ences therein) which, if unrecognized, can result ina failed oil–source correlation effort. Less well-rec-ognized compositional changes that often go com-pletely unnoticed can occur as a result of exchangebetween oil and carrier/reservoir rock at the molec-ular level (Philp and Gilbert, 1982; Hughes andDzou, 1995; Curiale, 2002). Pang et al. (2003) haveshown that less than 5–10% of a molecular contri-bution from the migration pathway or reservoircan alter biomarker compositions significantly, aphenomenon which would easily lead to erroneousoil–source rock correlations. This migration–con-tamination process is easily interpretable for oilswhich have migrated into a thermally immature partof the stratigraphic section (e.g., Fig. 5), and is seenat the extreme when recent organic matter is mixedwith migrated crude oil at the seawater–sediment

-36

-35

-34

-33

-32

-31

-30

-29

-28

-27

20 21 22 23 24 25 26 27 28

N-Alkane Carbon Number

Car

bo

n Is

oto

pe

Rat

io (

o/o

o)

2050 m

2210 m

2400 m

2610 m

Dashed Lines:

4 source rock samples from an 800m lacustrine section

composite of all 4 source rock samples

dominant lacustrine-sourced oil in the basin

Fig. 4. Distribution of carbon isotope ratio values for n-C20 through n-C28 in four lacustrine source rock samples (dashed curves; sampledepths are shown in the inset), one composite of data from these four samples (solid curve with circle symbols), and a lacustrine-sourcedcrude oil (solid curve with diamond symbols). All samples are from the Camamu Basin, offshore Brazil. Additional details are in theaccompanying text. The figure is modified extensively after a similar figure in Curiale and Sperry (2000). Note that although no singlesource rock extract correlates well with the oil, the composite curve of all four source samples is an excellent match to the oil.

J.A. Curiale / Organic Geochemistry 39 (2008) 1150–1161 1157

interface. In contrast, the phenomenon will often beentirely unrecognizable when oils reside in a maturepart of the stratigraphic section.

Adding to the problem of deconvoluting theeffect of migration–contamination on a reservoiredoil is the likelihood that this molecular exchange willvary in type and extent as a function of lateral andtemporal differences in the syndepositional organiccontent of the reservoir rock. If these differencesare significant, and occur at high frequency in therock, then oils tested from individual portions ofthe same reservoir in the same field can appear com-positionally different from one another (Hughes andDzou, 1995; Curiale and Bromley, 1996a). The suc-cess of oil–source rock correlations in such a settingcan deteriorate even further in situations wherephase changes during migration tend to fractionatethe product compositionally (Gussow, 1954;Thompson, 1988; Curiale and Bromley, 1996b); anexample is presented in Fig. 6. These migration-

related effects are additive through time and space,and provide a complicated scenario: indeed, migra-tion-induced effects have the potential to alter anoil’s composition to such an extent that an accurateoil–source rock correlation might be impossible(Curiale, 2002).

6. Recommended approach

The previous sections have provided examples offour situations where oil–source rock correlationscan be compromised: where laboratory rock extrac-tion techniques differ dramatically from natural oilexpulsion; where multiple source units and maturitylevels occur in the basin, resulting in oil mixingwithin the fetch area; where lateral and temporalorganic variability within a source unit (i.e., organicfacies difference) is substantial; and where migra-tion-induced compositional changes in an oil areappreciable and vary at small reservoir scales. These

A (oil)

B (coal extract)

20S

20R

Fig. 5. m/z 217 mass chromatograms of (A) Cook Inlet (Alaska)Sunfish-1 oil, and (B) a solvent-extract of a coal penetrated by theSunfish-1 well. The prominent peaks in the coal extract are 20S-5a,14a,17a-ethylcholestane (early eluting) and 20R-5a,14a,17a-ethylcholestane, and the low 20S/20R epimer ratio suggeststhermal immaturity. This low ratio has been overprinted on thesterane distribution of the Sunfish-1 oil, yielding a 20S/20Repimer ratio which implies a maturity level below that conven-tionally considered necessary to generated crude oil (figuremodified after Hughes and Dzou, 1995). The modified low ratioin this oil is representative of many molecular changes which canresult from migration–contamination, and which make correla-tion of the oil to its source rock particularly difficult.

Fig. 6. ‘‘Light-end bias” of n-alkanes in oils and condensates ofthe Vermilion-39 field, offshore Louisiana, USA, relative toreservoir depth. All samples shown are source-invariant, andwere expelled at similar maturity. Shallow oils are progressivelyenriched in lighter n-alkanes due to migration–fractionation.Post-sourcing compositional changes such as these, if unrecog-nized, can compromise oil–source rock correlations. Figuremodified from (and full details are present in) Curiale andBromley (1996b); four outliers, shown in the original paper, arenot shown here.

1158 J.A. Curiale / Organic Geochemistry 39 (2008) 1150–1161

difficulties are inherent in the complexity of the nat-ural system, and will probably always remain too

complex to unravel completely. However, specificapproaches can be used to minimize associatedproblems and provide a framework for reducingexploration risk in both frontier and maturebasins. These recommendations are summarizedbelow.

1. Select representative samples.a. Analyze at least one oil sample from each

fetch volume in the study area, and establishviable oil–oil correlations and family group-ings. If multiple source units are anticipatedin a single fetch volume, the number of oilsamples in this volume should at least equalthe number of mature source units in thevolume.

b. Analyze at least one rock sample from eachorganic facies in each source unit, preferablyat a maturity level corresponding to a trans-formation ratio between 10% and 50% (note:all rock samples must be pre-determined tobe source rocks, using standard criteria).

2. Establish the minimum necessary source rockvertical sampling resolution.a. Analyze sufficient samples within each source

unit to confirm that temporal variabilitycauses vertical source compositional differ-ences which are less than the compositionaldifferences observed between any two oil fam-ilies in the study area.

3. Assess the extent of migration-induced changes.a. If reservoirs occur in thermally immature sec-

tions, analyze the reservoired oils to confirmthat ‘‘immature biomarkers” (e.g., sterenes,hopenes; ring-unsaturated terrigenous trit-erpenoids) are absent or in low con-centration.

b. If long-distance vertical migration pathwaysare evident (e.g., growth faults), use physicalparameters of the fluids (e.g., API, GOR) toestablish that oil families are, indeed,source-defined, and not influenced by migra-tion–fractionation.

4. Support all oil–source correlations with 4dmodeling.a. Construct a 4d model of the study area to con-

firm that reasonable migration pathways con-nect, through both space and time, eachproposed source with each of its correlatedoil samples.

5. Iterate correlation results with ongoing explora-tion results.

J.A. Curiale / Organic Geochemistry 39 (2008) 1150–1161 1159

a. Maintain the 4d model and the oil/rock dat-abases ‘‘live”, integrating each additionaldata point and iterating the oil–source corre-lation conclusions and the modelaccordingly.

These five recommendations represent optimalconditions for an oil–source rock correlationattempt. Of course, in practice such ideal conditionsare rarely achieved. However, setting these recom-mendations as objectives provides a baseline levelfor assessing the reliability which decision makersshould place in oil–source rock correlations. Thelack of oil samples from a specific fetch volume;the lack of source rock samples from within theoil window; the absence of knowledge about verticalfacies changes or migration-induced changes – all ofthese situations compromise the success of an oil–source rock correlation, and must be factored intothe exploration risk assigned to a specific basin orregion.

In summary, then, the success of an oil–sourcerock correlation will be proportional to our successin achieving the objectives framed in these five rec-ommendations. To the extent that we can buildrobust datasets which contain proper sample types,sufficient sampling density, and geological under-standing, and to the extent that we support our con-clusions with holistic models, oil–source rockcorrelations will be successful. More importantly,because achieving these objectives will make eachcorrelation defensible, we will successfully reducethe exploration risk in both frontier and maturesettings.

7. Parting comment – oil–source rock correlations

using only oil

Traditional oil–source rock correlations, as dis-cussed throughout this paper, establish composi-tional and geological comparisons between oilsamples and rock samples. However, increasedunderstanding of petroleum systems and, in partic-ular, detailed knowledge of the molecular and isoto-pic composition of crude oils and the organic matterin their source rocks has led to a non-traditionalapproach to oil–source rock correlation which I willrefer to here as the inversion approach.

Inversion involves the use of detailed composi-tional information about a crude oil to infer thecharacteristics of its source(s). Inversion methodsare already in regular use to deduce, for example,

source rock maturity at the time of expulsion, orin-reservoir biodegradation level, and similar tech-niques for successfully ‘‘inverting” an oil’s composi-tion to assess its source character will undoubtedlyproliferate. Source rock age, environment of deposi-tion and lithology are now routinely ‘‘predicted”

from oil composition, based upon pre-establishedbasinal and regional source rock learning sets. Thereliability of this oil-only approach currently canbe, and should be, subject to serious critical apprai-sal. Nevertheless, progressively improving datasetsand incorporation of geological information intobasin models will gradually reduce the error inher-ent in this predictive approach. In summary, contin-ued understanding of the effect of source rockcharacteristics on the molecular and isotopic com-position of crude oil has the potential to minimizethe need for source rock analysis as a correlationalapproach, and it is likely that the use of inversionmethods will increase significantly in the years tocome.

Acknowledgements

Numerous Unocal and Chevron co-workers haveprovided ideas and correlation examples which haveaffected my thinking about oil–source rock correla-tions, and their help and advice are appreciated. Mythanks go to both of these organizations for provid-ing the regional petroleum systems overviews andextensive databases and analytical capabilities nec-essary to evaluate petroleum source rocks and theirassociated oils and gases in detail, and for providingapproval to present and publish this work. I alsoappreciate comments from R. Patience, B. Katz,an anonymous reviewer, and especially from M.Fowler, on an early version of the manuscript.

Associate Editor—Erdem Idiz

References

Aplin, A.C., Bishop, A.N., Clayton, C.J., Kearsley, A.T.,Mossman, J.R. Patience, R.L., Rees, A.W.G., Rowland,S.J., 1992. A lamina-scale geochemical and sedimentologicalstudy of sediments from the Peru Margin (Site 680, ODP Leg112). In: Summerhays, C.P., Prell, W.L., Emeis, K.C. (Eds.),Upwelling Systems: Evolution since the Early Miocene.Geological Society Special Publication No. 64, pp. 131–149.

Bailey, N.J., Burwood, R., Harriman, G.E., 1990. Application ofpyrolysate carbon isotope and biomarker technology toorganofacies definition and oil correlation problems in NorthSea basins. Organic Geochemistry 16, 1157–1172.

1160 J.A. Curiale / Organic Geochemistry 39 (2008) 1150–1161

Barker, C.E., Pawlewicz, M., Cobabe, E.A., 2001. Deposition ofsedimentary organic matter in blacks shale facies indicated bythe geochemistry and petrography of high-resolution samples,Blake Nose, western North Atlantic. In: Western NorthAtlantic Palaeogene and Cretaceous Palaeoceanography.Geological Society, London, Special Publication 183, pp.49–72.

Burwood, R., De Witte, S.M., Mycke, B., Paulet, J., 1995.Petroleum geochemical characterization of the Lower CongoCoastal Basin Bucomazi formation. In: PetroleumSource Rocks. Springer-Verlag, New York, pp. 235–263.

Chen, Jianping, Liang, Digang, Wang, Xulong, Zhong, Ning-ning, Song, Fuqing, Deng, Chunping, Shi, Xinpu, Tao, Jin,Xiang, Shuzheng, 2003a. Mixed oils derived from multiplesource rocks in the Cainan oilfield, Junggar Basin, NorthwestChina; Part I: genetic potential of source rocks, features ofbiomarkers and oil sources of typical crude oils. OrganicGeochemistry 34, 889–909.

Chen, Jianping, Deng, Chunping, Digang, Liang, Wang, Xulong,Zhong, Ningning, Song, Fuqing, Shi, Xinpu, Tao, Jin, Xiang,Shuzheng, 2003b. Mixed oils derived from multiple sourcerocks in the Cainan oilfield, Junggar Basin, Northwest China;Part II, Artificial mixing experiments on typical crude oils andquantitative oil–source correlation. Organic Geochemistry 34,911–930.

Curiale, J.A., 1993. Oil to source rock correlation – concepts andcase studies. In: Engel, M., Macko, S. (Eds.), OrganicGeochemistry, Topics in Geobiology, vol. 11. Plenum Press,New York, pp. 473–490.

Curiale J.A., 1994a. Correlation of oils and source rocks – aconceptual and historical perspective. In: Magoon, L.B.,Dow, W.G. (Eds.), The Petroleum System – From Source toTrap, vol. 60, American Association of Petroleum GeologistsMemoir, pp. 251–260.

Curiale, J.A., 1994b. High-resolution organic record of BridgeCreek deposition, northwest New Mexico. Organic Geochem-istry 21, 489–507.

Curiale, J.A., 2002. A review of the occurrences and causes ofmigration–contamination in crude oil. Organic Geochemistry33, 1389–1400.

Curiale, J.A., Bromley, B.W., 1996a. The Migration of petroleuminto Vermilion 14 Field, Gulf Coast, USA – molecularevidence. Organic Geochemistry 24, 563–579.

Curiale, J.A., Bromley, B.W., 1996b. Migration-induced compo-sitional changes in oils and condensates of a single field.Organic Geochemistry 24, 1097–1113.

Curiale J.A., Sperry, S.W., 2000. An isotope-based oil–sourcerock correlation in the Camamu Basin, offshore Brazil. In:Proceedings Volume of the 6th Latin American Congress onOrganic Geochemistry. Revista Latino Americana de Geo-quimica Organica, vol. 4, pp. 51–64.

Dow, W.G., 1974. Application of oil-correlation and source-rockdata to exploration in Williston Basin. American Associationof Petroleum Geologists Bulletin 58, 1253–1262.

Dzou, L.I., Holba, A.G., Ramon, J.C., Moldowan, J.M.,Zinniker, D., 1999. Application of new diterpane biomarkersto source, biodegradation and mixing effects on CentralLlanos Basin oils, Colombia. Organic Geochemistry 30, 515–534.

Eglinton, G., Murphy, M.T.J., 1969. Organic Geochemistry:Methods and Results. Springer-Verlag, New York.

Gibling, M.R., Haraluck, M., Tantisukrit, C., Thanasuthipitak,T., Uttamo, W., 1985. Oil shale sedimentology and geochem-istry in Cenozoic Mae Sot Basin, Thailand. AmericanAssociation of Petroleum Geologists Bulletin 69, 767–780.

Gussow, W.C., 1954. Differential entrapment of oil and gas – afundamental principle. American Association of PetroleumGeologists Bulletin 38, 816–853.

Huc, A.Y., Lallier-Verges, E., Bertrand, P., Carpentier, B.,Hollander, D.J., 1992. Organic matter response to change ofdepositional environment in Kimmeridgian shales, Dorset,UK. In: Organic Matter: Productivity, Accumulation, andPreservation in Recent and Ancient Sediments. ColumbiaUniversity Press, New York, pp. 469–486.

Hughes, W.B., Dzou, L.I.P., 1995. Reservoir overprinting ofcrude oils. Organic Geochemistry 23, 905–914.

Hunt, J.M., Stewart, F., Dickey, P.A., 1954. Origin of hydro-carbons of Uinta Basin, Utah. American Association ofPetroleum Geologists Bulletin 38, 1671–1698.

Jones, T.P., Fortier, S.M., Mosbrugger, V., Roessler, J., Ute-scher, T., Ashraf, A.R., 1997. 13C/12C ratio double cyclicity ina Miocene browncoal: isotopic signals and orbital forcing.Terra Nova 9, 19–23.

Keller, M.A., Macquaker, J.H.S., 2001. High resolution analysisof petroleum source potential and lithofacies of LowerCretaceous mudstone core (Pebble Shale Unit and GRZ ofHue Shale), Mikkelsen Bay State #1 Well, North SlopeAlaska. In: Houseknecht, D.W. (Ed.), NPRA Core Work-shop: Petroleum Plays and Systems in the National PetroleumReserve, Alaska. Society of Economic Paleontologists andMineralogists, SEPM Core Workshop, vol. 21, pp. 37–56.

Koel, M., Lyovin, S., Bondar, Y., 2000. Supercritical carbondioxide extraction of Estonian oil shael. Goryuchiye Slantsy[Oil Shale] 17, 225–232.

Larter, S.R., Huang, H., Adams, J., Bennett, B., Jokanola, O.,Oldenburg, T., Jones, M., Head, I., Riediger, C., Fowler, M.,2006. The controls on the composition of biodegraded oils inthe deep subsurface; Part II, Geological controls on subsur-face biodegradation fluxes and constraints on reservoir–fluidproperty prediction. American Association of PetroleumGeologists Bulletin 90, 921–938.

Lewan, M.D., 1985Evaluation of Petroleum Generation byHydrous Pyrolysis Experimentation, vol. 315. PhilosophicalTransactions of the Royal Society, London, pp. 123–134.

Leythaeuser, D., Mackenzie, A., Schaefer, R.G., Bjoroy, M.,1984. A novel approach for recognition and quantification ofhydrocarbon migration effects in shale-sandstone sequences.American Association of Petroleum Geologists Bulletin 68,196–219.

Li, H., Li, P., Miao, D., Shen, Z., Zhou, G., 1996. Supercriticalcarbon dioxide extraction of organic matter from petroleumsource rocks and its implications. Chinese Journal ofGeochemistry 15, 324–330.

Lin, R., Saller, A., Dunham, J., Teas, P., Kacewicz, M., Curiale,J.A., Decker, J., 2005. Source, generation, migration andcritical controls on oil vs. gas in the deepwater Kuteipetroleum system. In: Proceedings 30th Annual Convention& Exhibition of the Indonesian Petroleum Association, pp.447–466.

Nijenhuis, I.A., Bosch, H.-J., Sinninghe Damste, J.S., Brumsack,H.-J., De Lange, G.J., 1999. Organic matter and trace elementrich sapropels and black shales: a geochemical comparison.Earth and Planetary Science Letters 169, 277–290.

J.A. Curiale / Organic Geochemistry 39 (2008) 1150–1161 1161

Obermayer, M., Osadetz, K.G., Fowler, M.G., Snowdon, L.R.,2000. Light hydrocarbon (gasoline range) parameter refine-ment of biomarker-based oil–oil correlation studies: anexample from Williston Basin. Organic Geochemistry 31,959–976.

Pang, X., Li, M., Li, S., Jin, Z., 2003. Geochemistry of petroleumsystems in the Niuzhuang South Slope of Bohai Bay Basin.Part 2: evidence for significant contribution of mature sourcerocks to ‘‘immature oils” in the Bamianhe field. OrganicGeochemistry 34, 931–950.

Peters, K.E., Moldowan, J.M., Driscole, A.R., Demaison, G.J.,1989. Origin of Beatrice oil by co-sourcing from Devonianand middle Jurassic source rocks, Inner Moray Firth, UnitedKingdom. American Association of Petroleum GeologistsBulletin 73, 454–471.

Peters, K.E., Clutson, M.J., Robertson, G., 1999. Mixed marineand lacustrine input to an oil-cemented sandstone brecciafrom Brora, Scotland. Organic Geochemistry 30, 237–248.

Philp, R.P., Gilbert, T.D., 1982. Unusual distribution of biolog-ical markers in Australian crude oil. Nature 299, 245–247.

Rooney, M.A., Vuletich, A.K., Griffith, C.E., 1998. Compound-specific isotope analysis as a tool for characterizing mixed oils:an example from the West Shetlands area. Organic Geo-chemistry 29, 241–254.

Seifert, W.K., Moldowan, J.M., Jones, R.W., 1979. Applicationof biological marker chemistry to petroleum exploration. In:

Proceedings of the Tenth World Petroleum Congress, Heydenand Son, London, pp. 425–440.

Shinn, E.A., Robbin, D.M., Claypool, G.E., 1984. Compactionof modern carbonate sediments: implications for generationand expulsion of hydrocarbons. In: Palacas, J.G. (Ed.),Petroleum Geochemistry and Source Rock Potential ofCarbonate Rocks American. Association of Petroleum Geol-ogists Studies in Geology #18, Tulsa, Oklahoma, pp. 197–204.

Thompson, K.F.M., 1988. Gas-condensate migration and oilfractionation in deltaic systems. Marine and PetroleumGeology 5, 237–246.

Tyson, R.V., 1995. Sedimentary organic matterOrganic Faciesand Palynofacies. Chapman & Hall, London, p. 615.

Waples, D.W., Curiale, J.A., 1999. Oil–oil and oil–source rockcorrelations. In: Beaumont, E.A., Foster, N.H. (Eds.),Exploring for Oil and Gas Traps. American Association ofPetroleum Geology, Tulsa, pp. 8-1–8-71 (Chapter 8).

Welte, D.H., 1965. Relationship between oil and source rock.American Association of Petroleum Geologists Bulletin 49,2246–2268.

Williams, J.A., 1974. Characterization of oil types in WillistonBasin. American Association of Petroleum Geologists Bulle-tin 58, 1243–1252.