Light hydrocarbon (gasoline range) parameter re nement of biomarker-based oil±oil correlation studies: an example from Williston Basin

Transcription

1 Organic Geochemistry 31 (2000) 959±976 Light hydrocarbon (gasoline range) parameter re nement of biomarker-based oil±oil correlation studies: an example from Williston Basin Mark Obermajer *, Kirk G. Osadetz, Martin G. Fowler, Lloyd R. Snowdon Geological Survey of Canada, rd Street NW, Calgary, AB T2L 2A7, Canada Received 28 January 2000; accepted 26 July 2000 (returned to author for revision 27 April 2000) Abstract We evaluated geochemical compositions of 189 crude oils produced from Paleozoic reservoirs across the Williston Basin. Emphasis is placed on compositional variations in the gasoline range (i-c 5 H 12 -n-c 8 H 18 ) to verify the biomarkerbased classi cation of oil families. The oils belong to four distinct compositional oil families Ð A, B, C and DÐ broadly con ned to speci c stratigraphic intervals. The unique character of each oil family, evident from their n-alkane and biomarker signatures, is supported by distinctive gasoline range characteristics in general, and C 7 (``Mango'') parameters in particular. An invariance in the K1 parameter among oils from a single compositional group is observed for most of the oils. The K1 ratio, although relatively constant within each suite of oils, is di erent for each oil family, clearly indicating their compositional distinction. Other Mango parameters (N2, P2, P3) show a similar re ection of the oil families. However, while C 7 parameters provide excellent evidence for distinct familial association of oils from families A, B and D, family C often overlaps with the latter two families, perhaps indicating greater genetic and source heterogeneity in the family C oils. Nevertheless, di erences in the gasoline range composition suggest that the existing biomarker-based classi cation of oil families can be more universally applied throughout the entire Williston Basin. Moreover, because the light hydrocarbon parameters prove very useful in re ning oil±oil correlations, routine gasoline range analysis shows good potential as a supplementary component in geochemical correlation of crude oils, especially when high levels of thermal maturity decrease the usefulness of biomarker compounds. # 2000 Elsevier Science Ltd. All rights reserved. Keywords: Light hydrocarbons; Isoheptanes; ``Mango'' parameters; Oil±oil correlation; Williston Basin 1. Introduction Although higher molecular weight biomarkers (C 20 ± C 40 ) are considered the best tools for oil±oil correlation studies because they provide much information regarding an oil and its source rock (Peters and Moldowan, 1993), these compounds are unstable under thermal stress and are often absent in high maturity oils/condensates (van Graas, 1990; ten Haven, 1996). In contrast, many lower molecular weight hydrocarbon compounds, * Corresponding author. address: mobermaj@nrcan.gc.ca (M. Obermajer). though more susceptible to biodegradation, typically comprise a persistent fraction of oils at high maturities. Benchmark studies (Thompson, 1983; Mango, 1990; BeMent et al., 1995; Halpern, 1995; ten Haven, 1996) have suggested that gasoline range hydrocarbons also carry useful information regarding genetic associations and alteration of oils. It has been documented that the light hydrocarbon ratios have applications for oil-oil correlation studies (Mango parameters, C 7 -based star diagrams), provide an indication of the temperature of oil expulsion from its source (2,3-/2,4-dimethylpentane ratio), and re ect the stage of thermal decomposition of oil (para n indices). The application of these light hydrocarbon analyses is advantageous, not only because /00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved. PII: S (00)

2 960 M. Obermajer et al. / Organic Geochemistry 31 (2000) 959±976 they may constitute the only compositional fraction available for analysis in oils/condensates generated late during catagenesis when sterane and terpane biomarkers are below detection limit, but also because such techniques are relatively rapid and inexpensive. Therefore, they show excellent potential and can be extremely practical for geochemical correlations of low-density crude oil/condensate fractions providing valuable information about di erences in source kerogen, depositional paeloenvironment, genetic a nities and petroleum alteration, data typically obtained through more advanced analyses of biomarkers. Moreover, it has been indicated that analyses of light hydrocarbons have application in oil-source correlation studies because the lighter (C 5 ±C 8 ) fraction of source rock kerogen can be evaluated through thermal extraction (Jarvie and Walker, 1997; Odden et al., 1998). The main objective of the present study was to examine a suite of 189 oils produced from the Red River, Winnipegosis, Bakken and Madison reservoirs (Middle Ordovician Ð Lower Mississippian), from both the American and Canadian portions of the Williston Basin, with emphasis placed on the composition of the gasoline range hydrocarbons (C 5 ±C 8 range). These light hydrocarbon parameters, C 7 in particular, are used to constrain the biomarker-based classi cation of oil families in the Williston Basin. Mango (1987, 1990) observed a unique invariance in the relative concentration of methylhexanes and dimethylpentanes in oil, indicating that the ratio of [2-methylhexane+2,3-dimethylpentane]/[3-methylhexane+2,4-dimethylpentane], the so called K1 parameter, is relatively constant and remains around unity (i.e. 1.0). A high consistency of this ratio within a large set of oils (2000) was interpreted by Mango (1987) as an argument against a speci c biological precursor for isoheptanes. Instead, a chemical steady-state kinetic process, with constant rates of product formation, was proposed as a mechanism for the generation of isoheptanes. However, when discussing some other parameters derived from his C 7 parentdaughter transformation scheme, Mango (1990) also indicated that distinctions between some of these parameters likely re ect di erences in kerogen type and kerogen structure. Therefore, oils generated from the same source kerogen (homologous oils) should have similar ratios of isoheptanes and dimethylcyclopentanes. This concept was further tested by ten Haven (1996) who, based on a smaller (500) but global set of oils, concluded that K1 ratios should be consistent within cogenetic suites of oils. Although con rming the remarkable invariance of isoheptanes, ten Haven (1996) documented that the K1 ratio is not always around 1.0, and can vary signi cantly between homologous series of oils. Interestingly, this variance makes the K1 ratio very useful for correlating oils because ``...if K1 would have been constant for oils world-wide, then there would have been no application in correlation studies...'' (ten Haven, 1996, p. 962). It was stressed, however, that the light hydrocarbon parameters should be used in conjunction with other, more conventional geochemical data. More recently, Wilhelms et al. (1999) indicated that kinetic fractionation model proposed by Mango (1990) was inconsistent with compound speci c isotopic composition of C 7 hydrocarbons. These authors, however, also pointed to a common precursor for most of the C 7 compounds. Following these studies, C 7 parameters have been applied successfully in the Williston Basin for grouping oils (Jarvie and Walker, 1997; Obermajer et al., 1998). In the present paper, a number of standard gasoline range hydrocarbon parameters are used not only to examine if they are universally applicable but also to validate and re ne the existing biomarker-based classi- cation of oil families in the Williston Basin (Osadetz et al., 1992), and to investigate if this classi cation is applicable throughout the entire Williston Basin. Reexamination of this classi cation based on our new analyses will allow a much better understanding of the petroleum systems in the Williston Basin, providing a framework for appraising the future hydrocarbon potential of this basin. 2. Paleozoic oils in Williston basin Ð an overview The Williston Basin, situated on the western Canadian Shield within the interior platform structural province (Fig. 1), is a sub-circular epicratonic, preservational basin lled with sedimentary rocks of predominantly marine origin. These sedimentary sequences range in age from Cambrian to Tertiary reaching a maximum thickness of 5 km near the center of the basin (Williston, North Dakota). The basin is a proli c petroleum province with numerous occurrences of oil documented throughout the Phanerozoic succession. Petroleum occurs in structural, stratigraphic and combined structural±stratigraphic traps that are often controlled by important epeirogenic basement structures such as Cedar Creek and Nesson anticlines (Clement, 1987; Gerhard et al., 1987 LeFever et al., 1987). A rst attempt to classify oils from the Williston Basin was made by Williams (1974) who recognized three main oil types (Table 1). Oils occurring predominantly in Ordovician and Silurian reservoirs were identi ed as type I and attributed to sources in Middle Ordovician Winnipeg shales. type II oils, broadly corresponding to Upper Devonian, Mississippian and Mesozoic reservoirs were inferred to have a Bakken Formation source. A third group consisted of oils restricted to Pennsylvanian reservoirs and categorized as type III, with sources attributed to the Tyler Formation. Subsequent studies documented that carbon and sulphur

3 M. Obermajer et al. / Organic Geochemistry 31 (2000) 959± Fig. 1. Map showing the location and main structural elements of the Williston Basin. Table 1 Generalized Williston Basin oil family classi cation schemes (modi ed from Osadetz et al., 1994) Williams, 1974 Zumberge, 1983; Leeheer and Zumberge, 1987 Osadetz et al., 1992, 1994 Source rocks Type III (Pennsylvanian oils) not studied Type II (Devonian, Mississippian & Mesozoic oils) Not studied Not studied Tyler Fm. (Pennsyl.) Group 2 (Mission Canyon oils) Not studied Group 4 (Nisku oils) Group 3 (Duperow oils) Type 1 Group 1 (Ordovician-Silurian oils) (Red River oils) Group 5 (Cambrian oil) Family E (Bakken oils) Family B (Bakken oils) Family C (Miss. & Jurassic oils) Family D (Winnipegosis oils) Family A (Red River oils) Not studied Exshaw/Bakken Fm. (U. Dev.-Miss.) Bakken Fm. (U.Dev.-Miss.) Lodgepole Fm. (L. Miss.) Winnipegosis Fm. (M.Dev.) Winnipeg Gr. (M. Ord.) and Bighorn Gr. (U.Ord.) unknown (?U.Cam.-Ord)

4 962 M. Obermajer et al. / Organic Geochemistry 31 (2000) 959±976 isotope compositions conformed to these families and, together with other data, showed that many pools were water washed or biodegraded (Bailey et al., 1973a,b; Thode, 1981; Thompson, 1983). Based on a smaller sample set, Zumberge (1983) and Leenheer and Zumberge (1987) classi ed crude oils from the American Williston Basin into ve oil groups/families (Table 1). Group 1, identical to Type 1, was recognized as unique to Ordovician petroleum systems in the American mid-continent (Longman and Palmer, 1987; Jacobson et al., 1988; Foster et al., 1989) and elsewhere (Fowler and Douglas, 1984; Foster et al., 1986; Ho mann et al., 1987). Group 2 oils had characteristics similar to type II, Bakken-sourced oils. No equivalent to type III was identi ed. Instead, three other groups included high and low maturity oils from Devonian pools (groups 3 and 4, respectively) and a low maturity oil from Middle Cambrian ± Lower Ordovician Deadwood Formation (group 5). A commonly accepted model of eastern Williston Basin petroleum systems, outlined initially by Brooks et al. (1987) and then revised and more comprehensively documented by Osadetz et al. (1992, 1994) in Canada, was followed by comparable petroleum systems described from the US portion of the Williston Basin (Price and LeFever, 1994; Osadetz et al., 1995). Using combinations of terpane, steroidal and normal alkane characteristics Osadetz et al. (1992) categorized oils from the Canadian Williston Basin into four families. Family A, commonly restricted to Upper Ordovician reservoirs, had distinctive n-alkane distributions low acyclic isoprenoid/n-alkane ratios and corresponded to type I± group 1 (Table 1) with sources in Upper Ordovician Bighorn Group kukersites (Osadetz et al., 1992; Osadetz and Snowdon, 1995), and not Winnipeg shales as previously postulated (Williams, 1974; Dow, 1974). Oils with low tricyclic/pentacyclic terpane ratios but lacking gasoline range and n-alkane characteristics of the family A, typically occurring below the Three Forks Group, were classi ed as family D. This family was further subdivided into two sub-families recognized by their distinctive stratigraphic occurrence and n-alkane/ acyclic isoprenoid composition (Osadetz et al., 1992). Oils from Winnipegosis pinnacle reefs, family D2, are distinguished from other oils occurring predominantly in Saskatchewan and Manitoba groups reservoirs, family D1. Family D2 oils were speci cally inferred to have source rocks in the Brightholme Member of the Winnipegosis Formation, while D1 oils were inferred to have sources in Devonian strata like, but not exclusive to, those found at the contact between the Upper and Lower members of the Winnipegosis Formation (Osadetz et al., 1992; Osadetz and Snowdon, 1995). Family D was not represented in the original study (Williams, 1974), but would likely correlate with group 3, 4 and possibly group 5 oils of Leenheer and Zumberge (1987) (Table 1). Two other families, B and C, distinguished from families A and D mainly based on terpane ratios (Osadetz et al., 1992), are found in Bakken Formation to Mannville Group reservoirs. Family B oils occur primarily in the Bakken Formation, while Family C oils are found primarily in the Mississippian Madison Group and Mesozoic strata. Both families are subdivisions of type II of Williams (1974) and group 2 of Leenheer and Zumberge (1987). Although a Bakken source was initially inferred for all these oils, it has been proposed that only family B oils are derived from Bakken Formation shales and family C oils are derived from Lodgepole Formation carbonates (Osadetz et al., 1992, 1994; Price and LeFever, 1994; Osadetz and Snowdon, 1995). Families B and C were then identi ed in American Williston Basin by Price and LeFever (1994) who con rmed the predominance of family C oils in the Mississippian subcrop play and the common restriction of family B oils to the Bakken Formation. More recent studies of Williston Basin petroleum systems indicated a possible existence of several petroleum sub-systems and numerous source rock intervals within Madison Group strata (Jarvie and Inden, 1997; Jarvie and Walker, 1997). Moreover, an up-to-date assessment of the Williston Basin petroleum systems provided by Jarvie (in press) documents a dominant Madison Group system with four proven and two hypothetic sub-systems, as well as functional secondary systems, such as Bakken-Lodgepole, Bakken, Duperow and Red River petroleum systems. There are oils from a few pools, distinguished by their stratigraphic occurrence and isotopic composition, that do not comply with the general classi cation of the Williston Basin oils. These include a Cambrian Deadwood Formation oil at Newporte Field (Leenheer and Zumberge, 1987; Fowler et al., 1998) and a Beaverlodge Silurian pool on the Nesson Anticline (Downey, 1996). Therefore, there is a possibility that a major, currently unrecognized petroleum system (or systems) operates in the lower Paleozoic strata across the Williston Basin. More recently, Obermajer et al. (1999) indicated that oils occurring in the Upper Devonian Birdbear Formation reservoirs in Saskatchewan have a distinctive geochemical composition and should be separated from family D Winnipegosis oils, with which they were formerly grouped (Osadetz et al., 1992). 3. Analytical techniques The gasoline range hydrocarbons (i-c 5 H 12 ±n-c 8 H 18 ) were analysed on a HP5890 gas chromatograph connected to an OI Analytical 4460 Sample Concentrator. A small amount of the whole crude oil was mixed with deactivated alumina and transferred to the sample concentrator. The gasoline fractions were then passed onto

5 M. Obermajer et al. / Organic Geochemistry 31 (2000) 959± the gas chromatograph equipped with a 60m DB-1 fused silica column. The initial temperature was held at 30 C for 10 min and then programmed to 45 C at a rate of 1 o C/min. The nal temperature was held for 25 min. The eluting hydrocarbons were detected using a ame ionization detector. An aliquot of the fraction boiling above 210 C was deasphalted by adding an excess of pentane (40 volumes) and then fractionated using open column liquid chromatography. Saturated hydrocarbons were analysed using gas chromatography (GC) and gas chromatography±mass spectrometry (GC±MS). A Varian 3700 FID gas chromatograph was used with a 30 m DB-1 column coated with OV-1 and helium as the mobile phase. The temperature was programmed from 50 to 280 C at a rate of 4 C/min and then held for 30 min at the nal temperature. The eluting compounds were detected and quantitatively determined using a hydrogen ame ionization detector. The resulting gasoline range (GRGC) and saturate fraction chromatograms (SFGC) Fig. 2. Generalized Paleozoic stratigraphy in the Williston Basin.

6 964 M. Obermajer et al. / Organic Geochemistry 31 (2000) 959±976 were integrated using Turbochrom software. ``Mango'' parameters were calculated using the normalized percentage peak area from GRGC's instead of the weight percentage abundance in the whole oil as originally applied by Mango (1987). Single ion monitoring GC±MS experiments were performed on a VG 70SQ mass spectrometer with a HP gas chromatograph attached directly to the ion source (30 m DB-5 fused silica column used for GC separation). The temperature, initially held at 100 C for 2 min, was programmed at 40 C/min to 18 C and at 4 C/min to 320 C, then held for 15 min at 320 C. The mass spectrometer was operated with a 70 ev ionization voltage, 300 ma lament emission current and interface temperature of 280 o C. The instrument was controlled by an Alpha Workstation using Opus software. Terpane and sterane ratios were calculated using m/z 191 and m/z 217 mass fragmentograms. 4. Results and discussion Most of the analyzed oils are para nic in nature as they contain a high proportion of hydrocarbons in fraction boiling above 210 C. In general, the combined amounts of hydrocarbon fractions are higher in samples collected from the pools located in the southern (US) portion of the Williston Basin, often reaching values of more than 95%. The proportion of hydrocarbons is typically lower in oils from the northern (Canadian) part of the Williston Basin, although in most of the family D and some of the family B oils this parameter is often greater than 90%. Lower values (<80%) observed in oils from families A and C are associated with higher aromaticity and characteristic saturates/aromatics ratios of less than 1.0. In contrast, saturates predominate over aromatics in oils containing a high proportion of hydrocarbons resulting in saturates/aromatic ratios of more than 1.0. Data from the present study not only demonstrate that the examined oils belong to four distinct oil categories corresponding to previously de ned oil families A, B, C and D (Osadetz et al., 1992; 1994) but also indicate that the previously de ned biomarker-based classi cation of oil families can be more universally applied in the Williston Basin Evidence of familial association based on higher molecular weight (C 13 ±C 40 ) compounds Representative saturate fraction gas chromatograms (SFGC) and mass fragmentograms are shown on Figs. 3,6 and 8 while selected geochemical characteristics are summarized in Table 2. The di erences in the distributions of normal alkanes in C 13 ±C 30 range in each of the previously de ned oil families are evident from their SFGCs (Fig. 3). Family A oils (majority from the Fig. 3. Representative 210 C+ saturate fraction gas chromatograms showing variations in the n-alkane pro les in the Williston Basin oils. Pr-pristane, Ph-phytane, 15, 20, 25-C 15,C 20,C 25 normal alkanes.

7 M. Obermajer et al. / Organic Geochemistry 31 (2000) 959± Fig. 4. Cross-plot of phytane/n-c 18 H 38 (Ph/nC 18 ) ratio versus pristane/n-c 17 H 36 (Pr/nC 17 ) ratio for the oil samples, showing typical elds incorporating most of the oils from each oil family (outlines are intended for visual approximation only). Black symbols denote two family B samples with Pr/nC 17 ratios of 1.22 and Fig. 5. Cross-plot of pristane/phytane ratio (Pr/Ph) versus Carbon Preference Index (CPI, C range, see Table 2 for ratio de nition) for the analyzed oil samples. The outlines, showing typical elds incorporating most of the oils from each oil family, are intended for visual approximation only. Black symbols denote one family A and one family C sample with CPI values of 1.98 and 2.19, respectively. Ordovician Red River/Yeoman reservoirs) are characterized by n-alkane pro les centered at C 13 ±C 17, very low concentration of acyclic isoprenoids (i.e. pristane and phytane) relative to n-alkanes (lowest pristane/n- C 17 H 36 and phytane/n-c 18 H 38 ratios Ð Fig. 4) and low relative abundance of C 20+ members, resulting in high middle/long chain n-alkane ratios, often greater than A strong predominance of odd to even normal alkanes within the C 14 ±C 20 range and high Carbon Preference Index values (average CPI of 1.59) are typical in these oils (Fig. 5). The absolute Pr/Ph, Pr/nC 17 and Ph/nC 18 ratios have to be carefully veri ed as there may be a bias due to very low concentrations of pristane and phytane (co-elution with other compounds during GC analysis). Regardless, these oils can be easily di erentiated from the other families based on their overall n- alkane pro le and CPI values. As previously noted by Osadetz et al. (1992), these characteristics are common to many Ordovician oils worldwide (Reed et al., 1986; Longman and Palmer, 1987; Fowler, 1992) indicating they originated from sources geochemically similar to those of the other Ordovician oils, with the algae Gloeocapsomorpha prisca as a main component of the source organic matter. The remaining groups of oils have broader n-alkane pro les with relatively higher concentrations of C 20+ members and acyclic isoprenoids. Oils from Family B are characterized by a smooth n-alkane distribution with a maximum in the C 13 ±C 17 range. Their pristane/ phytane ratios are usually greater than 1.0, similar to family A, but the concentration of acyclic isoprenoids relative to n-alkanes is much higher than in family A oils, allowing for obvious separation (Fig. 4). Moreover, the concentration of long-chain hydrocarbons is also higher in these oils when compared with family A while the CPI values are lower (Fig. 5). Families C and D are characterized by predominance of phytane over pristane (stronger and more consistent in family C) and Pr/Ph ratios of less than 1.0 (Fig. 3, Table 2). The concentration of phytane relative to C 18 n-alkane is also higher in these groups as compared with families A and B. However, as shown on Fig. 4, there is some overlap in Pr/ nc 17 and Ph/nC 18 ratios among oils from families C and D. The low Pr/Ph ratios and high abundances of phytane indicate that the source kerogen was deposited under restricted, elevated-salinity conditions. Accordingly, Family C oils have been correlated to sources deposited in distal anoxic settings of a broad carbonate ramp (Lodgepole Fm.) while Family D oils have been related to sources from ``o -reef'' starved basins settings (Winnipegosis Fm.) (Osadetz et al., 1992, Osadetz and Snowdon, 1995). Both oil families have comparable concentrations of long-chain hydrocarbons and cannot be easily resolved from each other based on their mid-/ long-chain hydrocarbon ratios (Table 2). Nevertheless, the SFGCs of the Winnipegosis-sourced family D oils have a di erent n-alkane envelope, typically displaying a bimodal distribution centered at C 15 ±C 17 and C 23 ±C 25 (Fig.3). Furthermore, there is a stronger odd carbon number preference in family D oils, with an average CPI ratio of 1.25 compared with 1.09 in the family C oils (Fig. 5, Table 2). Therefore, SFGC analysis provides good evidence for the existence of four compositional oil families.

8 966 M. Obermajer et al. / Organic Geochemistry 31 (2000) 959±976 The geochemical di erences within each family of oils were originally shown by the distributions of steranes and terpanes. The oils are easily categorized based on the distribution of extended hopanes and C 34 or C 35 prominence. The family A oils are characterized by a smooth extended hopane pro le with a steady decrease in the concentration of C 31+ homologues with increasing carbon number, although some samples have a minor predominance of the C 34 member (Fig. 7). A more distinctive C 34 hopane prominence is readily observable in the Winnipegosis oils (Figs. 6 and 7). In contrast, most family C oils show a C 35 hopane prominence, a characteristic not noticeable in oils from other families of eastern Williston Basin. Moreover, these oils also have the highest relative abundance of C 29 norhopane and relatively high concentration of C 23 tricyclic terpanes, and therefore, are easily distinguishable as a separate family (Figs. 9 and 10). Higher abundance of C 23 tricyclic terpane is also seen in Family B oils (Fig. 10), but these oils also display relatively higher concentration of C 24 tetracyclic terpane compared with family C oils. Perhaps a more apparent di erence in the terpane ngerprints of these two families is the lack of any homohopane Fig. 7. Biomarker cross-plot showing C 35 and C 34 homohopane prominence in the investigated oil samples. The outlines, showing typical elds incorporating most of the oils from each oil family, are intended for visual approximation only. 33h-17(H), 21(H) 20S+20R- C 33 homohopane, 34h-17(H), 21(H) 20S+20R- C 34 homohopane, 35h-17(H), 21(H) 20S+20R- C 35 homohopane. Fig. 6. Representative m/z 191 mass fragmentograms of the saturate fraction showing the distribution of terpanes in the analyzed Williston Basin oils. 23-C 23 tricyclic terpane, Ts-18(H)-trisnorhopane, 30-hopane, G-gammacerane, 34-C 34 homohopane.

9 M. Obermajer et al. / Organic Geochemistry 31 (2000) 959± prominence in family B oils (Fig. 7). Similarly, C 23 tricyclic and C 24 tetracyclic terpane compounds are in low relative concentrations in oils from the other two families (A and D). There is a predominance of 17(H)-trisnorhopane (Tm) over 18(H)-trisnorhopane (Ts) in most of the analyzed oils with the exception of Family B oils in which the relative concentration of these two compounds is quite variable, with a majority of samples having Ts/Tm ratios greater than 1.0. There are some di erences in the normalized relative abundance of C 27 :C 28 :C 29 regular steranes. While the proportion of the C 28 member is the lowest and generally similar in all oil families, the proportion of C 27 and C 29 members is more variable. In the family B oils, the C 29 sterane occurs in same concentration as C 27 sterane (Table 2). In contrast, the relative abundance of C 29 sterane are increasingly greater in the remaining families, reaching more than 50% in oil families D and A (average of 51% and 58%, respectively). The proportion of C 27 :C 28 :C 29 regular steranes is considered to be a highly speci c correlation index (Peters and Moldowan, 1993, pp. 182±186), therefore providing further support for classifying studied oil samples into compositionaly distinct oil categories. Moreover, m/z 217 mass fragmentograms show prominent diasterane peaks (Fig. 8). Both C 29 and C 27 diasteranes are present in high concentrations relative to regular steranes especially in oil families A and B (Fig. 9). However, the B oils show a higher relative abundance of C 21 regular sterane (pregnane) and the highest C 21 /C 29 regular sterane ratio amongst all analyzed samples making distinction of the oil family B quite apparent (Table 2, Fig. 10). Some of the family A oils also display higher concentrations of C 21 regular sterane, which perhaps could be related to their higher thermal maturity (slightly higher C 29 regular sterane isomerization ratios) as the majority of oils from this family have C 21 /C 29 regular sterane ratio of less than 1.0. The epimerization ratios of C 29 regular steranes are quite variable and inconclusive with respect to determining maturity of oils. The C 29 abb/(aaa+abb) regular sterane isomerization ratio is the highest in oil families B and C, often approaching equilibrium values. Interestingly, this trend is not parallelled by the C 29 S/(S+R) isomerization ratio which is the highest in oil families A and D. These variable ratios are not always consistent with maturity data derived from the gasoline range parameters and the distribution of n-alkanes and Fig. 8. Representative m/z 217 mass fragmentograms of the saturate fraction showing the distribution of steranes in the Williston Basin oils. C21- pregnane, C27dia-17(H), 13(H), 17(H) 20S- C 27 diasterane, C29±C 29 regular sterane.

10 968 M. Obermajer et al. / Organic Geochemistry 31 (2000) 959±976 Fig. 9. Biomarker cross-plot of C 29 norhopane/c 30 hopane ratio (29h/30h) versus C 27 diasterane/c 27 regular sterane ratio (dia/reg) for the Williston Basin oil samples (see Table 2 for ratio de nitions). The outlines, showing typical elds incorporating most of the oils from each oil family, are intended for visual approximation only. Fig. 10. Biomarker cross-plot of C 23 tricyclic terpane/c 30 hopane ratio (23h/30h) versus C 21 /C 29 regular sterane ratio (21/ 29s) for the Williston Basin oil samples. Oil family B samples with 21/29s ratios greater than 2.35 (11 samples) are not plotted. The outlines, showing typical elds incorporating most of the oils from each oil family, are intended for visual approximation only. acyclic isoprenoids, what indicates a possible source control on the concentration of C 29 regular sterane and hence the di erent genetic character of these oils Compositional di erences among gasoline range (C 5 ±C 8 ) hydrocarbons Representative gasoline range gas chromatograms (GRGC) are shown on Fig. 11 while selected gasoline range characteristics of the investigated oils are listed in Table 2 and shown on Figs. 12±17. As there is commonly a variable overlapping in light hydrocarbon parameters between oil family C and other oil families, data for family C oils are typically shown on separate cross-plots ( gures ``b''). The relative abundances of the normal heptane and isoheptane, as well as the Para n Indices (Thompson, 1983), provide initial evidence for grouping the oils into separate categories (Figs. 12 and 13). However, while the distinction of oil families A, B and D is evident based on these characteristics, Family C is not clearly distinguishable. The family A oils have the highest concentration of n-heptane relative to isoheptane (Fig. 12a), and therefore, the highest and most variable n-heptane/isoheptane ratio (nc 7 /bc 7 Ð see Table 2 for ratio de nition) ranging from 1 to 10 (average of 5.19). The same parameter shows less variation in the remaining oils which have similar concentrations of these two compounds with their ratio close to 1.0. There is some overlap, but in general relatively good separation between oil families B and D (Fig. 12a). In contrast, it is di cult to clearly distinguish family C as most of these samples show similar proportions of heptane and isoheptane to family B, typically plotting within its eld (Fig. 12b). Similarly, both PI I and PI II values are generally lowest in the family B and C oils, with some overlap with family D oil samples (Fig. 13). The high values obtained for the Family A oils (Heptane Values greater than 30.0 and Isoheptane Values often greater than 1.0) easily classify these oils into a distinct group. Such variable Para n Indices would suggests that the family A oils are generally of the highest and families B and C of the lowest maturity. Accordingly, using the criteria of Thompson (1983) the A oils could be interpreted as supermature and the remaining groups as normal mature. Although the degree of para nicity can be useful in estimating thermal maturity such an assessment has to be constrained with other maturity indicators. High Para n Indices may also result from the retention of already generated hydrocarbons in source rocks prior to expulsion or extended presence of oil in the reservoirs (Thompson, 1983). However, as the concentration of cycloalkanes in oil depends not only on temperature and pressure but also on kerogen structure, it is more likely that observed low concentrations of cycloalkanes have resulted from natural variations in source organic facies due to the highly aliphatic, G. prisca-derived (Ordovician) source kerogen for the family A oils. The greater variability in Para n Indices observed within the family C and D oils likely re ects a more complex nature of source organic facies and the e ects of the restricted paleodepositional environment of their source rocks. There is also some variability in the concentrations of low molecular weight aromatic hydrocarbons such as benzene and toluene. The concentrations of toluene

11 M. Obermajer et al. / Organic Geochemistry 31 (2000) 959± Table 2 Selected geochemical characteristics of the Williston Basin oils analysed in the present study Family A total 58 spls Family B total 31 spls Family C total 63 spls Family D total 37 spls no. a mean st. dev. b no. mean st.dev. no. mean st.dev. no. mean st.dev. N-alkanes/ Pr/Ph c isoprenoids d Pr/nC e Pn/nC CPI f mc/lc g Sterane/terpane S/(S+R) h biomarkers = dia/reg j :28:29 k 28:14:58 39:20:41 38:15:47 29:20:51 21/29s l Ts/Tm m h/30h n h/30h o Gasoline range hydrocarbons P1 I p Mango parameters PI II q r nc s bc K1 t P2 u P3 v N2 w a no. Ð Number of analyses used. b st.dev. Ð standard deviation. c Pristane/phytane ratio. d Pristane/n-C 17 ratio. e Phytane/n-C 18 ratio. f Carbon Preference Index=1=2 C 15 C 17 C 19 = C 14 C 16 C 18 Š C 15 C 17 C 10 = C 16 C 18 C 20 Š. g Medium-/long-chain n-alkanes= C 15 C 16 C 17 C 18 C 19 = C 21 C 22 C 23 C 24 C 25. h 5a(H),14a(H),17a(H) 20S/(20S+20R)- C 29 sterane. i 5a(H),14b(H),17b(H)/[5a(H),14a(H),17a(H)+5a(H),14b(H),17b(H)] (20R+20S)- C 29 sterane. j 10a(H),13b(H),17a(H) 20S- C 27 diasterane/5a(h),14a(h),17a(h) 20R- C 27 sterane. k Normalized relative abundance of C 27,C 28 and C 29 regular steranes based on abb isomers. l Pregnane/5a(H),14a(H),17a(H) 20R- C 29 sterane. m 18a(H)-trisnorhopane/17a(H)-trisnorhopane. n 17a(H)-norhopane/17a(H)-hopane. o C 23 tricyclic terpane/17a(h)-hopane. p Para n Index I (Isoheptane Value- Thompson, 1983). q Para n Index II (Heptane Value- Thompson, 1983). r n-heptane*100/ of compounds eluting between 2-methylhexane and 2,2-dimethylhexane. s (2-methylhexane+2,3-dimethylpentane+3-methylhexane)*100/ of compounds eluting between 2-methylhexane and 2,2-dimethylhexane t K1=(2-methylhexane+2,3-dimethylpentane)/(3-methylhexane+2,4-dimethylpentane). u P2=2-methylhexane+3-methylhexane. v P3=dimethylpentane (2,2- +2,4- +3,3- +2,3-). w N2=dimethylcyclopentane (1,1- +1,c3- +1,t3-). relative to methylcyclohexane and benzene relative to 3- methylpentane, are generally the lowest in the B oils, especially in the Canadian samples, probably re ecting water washing and longer secondary migration pathways. Except for a few samples, benzene and toluene are present in higher relative concentrations in the remaining oil families, with family C oils typically showing the highest amounts of these compounds.

12 970 M. Obermajer et al. / Organic Geochemistry 31 (2000) 959±976 Fig. 11. Representative gasoline range gas chromatograms (i-c 5 H 12 -n-c 8 H 18 ) of the analyzed Williston Basin oils. 1-2,2-dimethylpentane, 2-2,4-dimethylpentane, 3-3,3-dimethylpentane, 4-2-methylhexane, 5-2,3-dimethylpentane, 6-1,1-dimethylcyclopentane, 7-3-methylhexane, 8-1,cis-3-dimethylcyclopentane, 9-1,trans-3-dimethylcyclopentane, 10- heptane, 11- methylcyclohexane, 12- toluene, 13- octane. Fig. 12. Relative concentrations of normal heptane (nc 7 ) and isoheptane (bc 7 ) (see Table 2 for ratio de nitions). The outlines are for visual approximation only. While there is some overlapping between oil families B and D (a), family C oils typically fall within family B eld (b). The separation of oils from the Williston Basin as de ned by light hydrocarbon parameters, is further re ned using selected C 7 ``Mango'' parameters (Mango, 1987, 1990). Previous studies elsewhere have demonstrated a possible general invariance in the K1 parameter (the ratio of [2- methylhexane+2,3- dimethylpentane]/[3- methylhexane+2,4-dimethylpentane]). In general, this is also observed for most of the studied samples. However,

13 M. Obermajer et al. / Organic Geochemistry 31 (2000) 959± Fig. 13. Para n Indices (Thompson, 1983) in the oils. The outlines are for visual approximation only. There is a good separation of oil family A and some overlapping between oil families D and B (a), while family C oils overlap with these both families (b). Fig. 14. Normalized % peak area of 3-methylhexane+2,4-dimethylpentane versus 2-methylhexane+2,3-dimethylpentane showing invariance of Mango's K1 parameter in the analyzed oil samples. The outlines are for visual approximation only. There is a good separation between oil families A, B and D (a) but some overlapping between oil families C and B (b). although these ratios appear relatively constant within each biomarker-de ned oil family they are not around unity (i.e. 1.0) as claimed by Mango (1987). Following the conclusions of ten Haven (1996) that K1 ratios would be expected to be consistent within co-genetic suites of oils, di erent values of K1 parameter determined for each oil family can be interpreted as verifying the genetic uniqueness of the oil families. Fig. 14a shows a good separation between families A, B and D providing an additional, strong evidence for such grouping. However, there is strong overlap between families C and B (Fig. 14b). This observation is compatible with indications of familial a nity from other light hydrocarbon parameters, such as those shown on Figs. 12 and 13. Moreover, the K1 ratios for oil families B and C are typically below 1.0 (on average 0.90 and 0.86, respectively Ð Table 2) which clearly contrasts with values greater than 1.0 determined for families A and D (on average 1.15 and 1.23, respectively). This range of K1 values corresponds to that reported by Schaefer and

14 972 M. Obermajer et al. / Organic Geochemistry 31 (2000) 959±976 Fig. 15. Normalized % peak area of Mango parameters P2+N2 versus P3 showing (a) good separation between oil families A, B and D, and (b) some overlapping between Family C and families D and B. The outlines are for visual approximation only. Fig. 16. Normalized % peak area of Mango parameters N2/P3 versus P2 showing (a) good separation between oil families A, B and D, and (b) some overlapping between family C and families D and B. The outlines are for visual approximation only. Littke (1988) who observed an increase in K1 values with increasing thermal maturity within a vitrinite re ectance range of 0.48±0.88%. However, as these results were obtained from thermal extracts of type II kerogen source rocks, where concentrations of light hydrocarbons is often a ected by di usion or early expulsion and migration, the e ect of temperature on K1 ratio in oils remains to be tested. Two other cross-plots similar to those used by ten Haven (1996) show familial a nity and variation in a similar manner. The data for families A, B and D plot in separate elds on the P2+N2 versus P3 and the N2/P3 versus P2 cross-plots (Figs. 15 and 16). This provides further support for classifying these oils into three separate oil families. In addition, these two correlations also indicate a similar di culty of distinguishing family C oils from the other oils. Moreover, it appears that the light hydrocarbon guidelines for distinguishing lacustrine from terrigenous/marine oils based on these parameters, as used by ten Haven (1996), are not universally applicable Discussion With the exception of family C oils, the light hydrocarbon data presented here not only support a biomarker-based classi cation, but also provide additional evidence for familial association of families A, B and D. All three families are characterized by di erent gasoline

15 M. Obermajer et al. / Organic Geochemistry 31 (2000) 959± Fig. 17. Ternary plot of C 7 hydrocarbons showing 5-carbon ring preference in Family B and 6-carbon ring preference in family C oils. Family B oils from northern part (N) of the basin show greater relative concentrations of dimethylcyclopentanes compared with those from more central basinal location (S). Tol+MCYC6- sum of toluene and methylcyclohexane, 2MC6+3MC6- sum of 2-methyl- and 3-methylhexane, DMC5+DMCYC5- sum of dimethylpentanes and dimethylcyclopentanes (2,2-dmp+2,4-dmp+3,3-dmp+2,3- dmp+1,1-dmcp+1,cis-3-dmcp+1,trans-3-dmcp+1,trans-2-dmcp). range data which commonly fall within relatively narrow compositional ranges. In contrast, the same light hydrocarbon parameters regularly show greater variability amongst family C oils. Moreover, a general similarity in the gasoline range composition of the family C samples to the B oils is quite intriguing considering the fact that not only the compositions of their higher molecular weight fractions but also the concentrations of sulfur in both groups are distinctly di erent (there is a much greater amount of sulfur present in family C oils). This observation is not in full agreement with the results of Jarvie and Walker (1997) who, based on C 6 and C 7 parameters, not only concluded that Madisonreservoired oils are distinctly di erent from Bakkenreservoired oils, but also indicated a positive correlation of both groups to sources occurring in the stratigraphic intervals of their reservoirs. Both families occur in stratigraphically separated compartments but upward migration of Bakken-sourced oils into Madison reservoirs with or without mixing with Madison-sourced oils, though relatively uncommon, is known to have occurred (Burrus et al., 1996). However, such cases are easily recognised using their intermediate or Bakken-like biomarker compositions (i.e. oils from Hummingbird and Ceylon elds in Saskatchewan, oils in Madison Group pools of the Nesson Anticline and Waulsortian mound oils in Stark County, North Dakota). The possibility of mixing between family B and C oils was recently assessed by Jarvie (in press) who concluded that Madison Group oils which show mixed signature could have received some contribution from Bakken sources, but more likely were generated from marly shale sources within Mission Canyon strata. On the other hand, oils with a family C biomarker signature are absent from pre-madison reservoirs (except for Hummingbird eld, Osadetz et al., 1992). Furthermore, some ``Mango'' parameters (P3, N2) in C oils are also similar to those of the family D oils (Figs. 15 and 16). One possible explanation for the greater variability within the C family oils and similarities in gasoline range fraction of those oils to pre-madison-sourced oils is the limitation of ``Mango'' parameters and the fact that this approach simply cannot be extended to the Madison Group oils. However, the light hydrocarbon fraction (boiling below 210 C) constitutes the bulk of the analyzed family B and family C oils, 70% and 65%, respectively. Moreover, the absolute concentrations of higher molecular weight biomarkers used as primary evidence of familial association are low in these oils, especially in the family B Ð Bakken-reservoired oils (M. Li, personal communication). Therefore, the gasoline range data deserve a more careful evaluation. It could be hypothesised that compositional inconsistency amongst family C oils is due to a greater

16 974 M. Obermajer et al. / Organic Geochemistry 31 (2000) 959±976 heterogeneity of their sources, such as multiple source rock bodies within the Madison Group interval, lateral organic facies variation or variable concentration of sulfur within the Lodgepole Formation. Studies from the American Williston Basin indicate that there are at least four proven and two hypothetic petroleum subsystems in the Mississippian strata with numerous oilprone source rocks in the Madison Group (Jarvie and Inden, 1997; Jarvie and Walker, 1997; Jarvie, in press). Most of the Family C oils analyzed during the present study come from the Canadian Williston Basin and no attempt has been made to subdivide them. The majority of these oils have been correlated to source rocks within the Lodgepole Formation (Osadetz et al., 1992, 1994), but most pools are in the overlying Charles and Mission Canyon formations. This indicates that cross-formational migration of oil from the underlying Lodgepole Formation must have occurred. Moreover, the Lodgepole Formation in the Canadian Williston Basin is of low thermal maturity with respect to hydrocarbon generation (Osadetz and Snowdon, 1995) and it has been documented that majority of these oils must have migrated considerable lateral distances from more central parts of the American Williston Basin (Osadetz et al., 1992, 1994; Burrus et al., 1996). However, although it is possible to envisage that the primary source signature have been altered during the secondary migration, one could expect that such long-distance (hundreds of kilometers), cross-stratal ow would result in a more homogenic composition of the oil. Modelling of oil expulsion and accumulation from the Bakken shales, which were long suspected to be the major source of oil found in the Madison Group reservoirs (Dow, 1974; Price et al., 1984), documented that most of the expelled Bakken oil remains largely dispersed within the drainage system of the Madison strata; the system that formed secondary migration fairway for the Lodgepole-sourced oil. Nevertheless, this oil appears to be immobile as the oil saturation rates are too low for a free ow to occur (Burrus et al., 1996). According to Muscio et al. (1994), Bakken kerogen shows enhanced tendency to generate light hydrocarbons, which generally corroborates the high API gravity character of the oils that form commercial accumulations in Bakken pools. If mixing of oils from Bakken and Lodgepole sources had been greater than previously estimated, the Bakken component could be assumed to be dominantly light hydrocarbons. Therefore, the Madison Group (family C) oils, despite having unique biomarker composition, would bear a strong resemblance in their gasoline fraction to the Bakken-sourced (family B) oils. However, there are some di erences in gasoline range composition between both groups. There is a 6-carbon ring preference in the Madison Group oils whereas the Bakken-reservoired oils are preferentially enriched in 5- carbon ring hydrocarbons (Fig. 17). These results are similar to those of Javie (in press) who indicated that 5- carbon ring preference may be a function of clay content in the Bakken Formation. Moreover, the family C oils are enriched in aromatic hydrocarbons, toluene and benzene, which are present in relatively lower concentration in family B oils, despite the fact that kerogen in Bakken Formation has been considered as having enhanced aromatic nature (Muscio et al., 1994; Muscio and Hors eld, 1996). The de ciency of both compounds in Bakken-reservoired oils could be due to water washing and long secondary migration (oils from northern Canadian locations generally have a lower relative concentration of toluene compared with those from more central parts of the basin, Fig. 17). Interestingly, most of the family C oils, despite having been subjected to long secondary migration (Osadetz et al., 1992, 1994; Burrus et al., 1996), appear to have relatively higher amounts of toluene and benzene. Jarvie (in press) indicated that the higher sulfur content in the Madison Group source kerogen might have played a role enhancing cyclization of straight-chain para ns leading to aromatization and greater production of toluene, although de ciency of free hydrogen may also result in preferential formation of toluene. It appears that the di erences in the light aromatic hydrocarbons might have been at least to some extent source-controlled. Therefore, the mixing hypothesis remains to be tested using other techniques such as analyses of compound-speci c isotopes, aromatic hydrocarbons, sulfur and other lower±molecular±weight fractions of both oil families and their sources. 5. Summary Detailed analyses of gasoline range (C 5 ±C 8 ) fraction in four Paleozoic oil families in the Williston Basin (families A, B, C and D) demonstrate that certain light hydrocarbon (``Mango'') parameters are useful in re ning oil-oil correlations. Familial compositional distinctions, originally de ned based on the distribution of saturated hydrocarbons in the C 13 ±C 30 range (odd/even n-alkane predominance, relative abundance of isoprenoids to n-alkanes) and terpane and sterane biomarkers, are extended to lighter hydrocarbon fractions. An invariance in the K1 ratio (the ratio of [2-methyl- hexane+2,3-dimethylpentane]/[3-methylhexane+2,4- dimethylpentane]) has been observed, but the absolute K1 values are di erent in each oil family, indicating their genetic distinction. Other parameters (P2, P3, N2), although providing additional evidence that the existing biomarker-based classi cation of oils rst de ned in Canadian Williston Basin is more universally applicable on the basin scale, also demonstrate considerable di culty with applying ``Mango'' parameters to distinguish the Madison Group oil family C from oils found in the underlying Bakken reservoirs (oil family B). Similarities