Correlation of weathered and unweathered oil samples using the carbon isotopic composition of individual components in the crude oil

Transcription

1 Correlation of weathered and unweathered oil samples using the carbon isotopic composition of individual components in the crude oil R.P. Philp, B. Smallwood & J. Allen School of Geology and Geophysics, University of Oklahoma, Norman, 7307P, Abstract Once a crude oil is spilled in the environment, changes to the composition of the spilled oil will occur. Evaporation and water washing cause the initial changes, followed by photo-oxidation and biodegradation. Correlations between spilled oils and suspected sources are normally based on techniques such as gas chromatography (GC) and gas chromatography-mass spectrometry (GCMS). Whilst generally successful, heavily weathered residues or refined products are often extremely difficult to correlate by GC or GCMS. In this paper results from a relatively new correlation technique, gas chromatography-isotope ratio mass spectrometry (GCIRMS) are discussed. GCIRMS permits the determination of the isotopic composition of individual components in complex mixtures. The isotopic composition of individual compounds is relatively unaffected by the effects of biodegradation. GCIRMS is particularly powerful for refined products which contain none of the biomarkers commonly used for correlation purposes. GCIRMS data can be utilized in conjunction with the conventional techniques and the end result is a powerful tool that has widespread applicability for monitoring and correlation studies. 1 Introduction In the past two or three decades there have been many studies concerned with hydrocarbon pollutants in aquatic environments [1-3]. Many studies have been directed toward tracing the source of contaminants, determining rates of degradation, and effects on the environment. Sources of hydrocarbon contaminants include: natural oil seeps; ship traffic; leaking storage tanks or pipelines. Identifying, quantifying, and monitoring, the fate of pollutants spilled in the

2 212 Oil and Hydrocarbon Spills II: Modelling, Analysis and Control environment are of primary importance in providing a better response to an hydrocarbon spill. Characterization of a spilled hydrocarbon product and the identification of its potential source generally relies on GC and GC-MS. For crude oils, correlation between pollutant and suspected source is based on the biomarkers fingerprints. GC and GC-MS data may be ambiguous, and sometimes unsuccessful, since oils, and refined products, released in the environment are quickly affected by weathering processes, such as evaporation, photooxidation, water washing, biodegradation. Evaporation quickly removes the more volatile hydrocarbons, water-washing the more water-soluble hydrocarbons, typically hydrocarbons below C,5, and some of the C,5+ aromatic compounds which are more water-soluble than paraffins [4-6]. Biodegradation will initially remove low carbon number n- paraffins, followed by branched paraffins and single-ring naphthenes and certain biomarkers [5]. Combined these weathering processes can modify the fingerprints used to correlate an oil with its source on the basis of GC-MS analysis. Whilst GC and GCMS generally work relatively well in terms of correlating weathered and unweathered crude oils, problems are exacerbated with refined products since they do not contain biomarkers. Correlation of a spilled refined product to its suspected source requires the use of discriminative parameters that are relatively insensitive to weathering processes. Bulk carbon isotopic compositions of refined products have been used to determine potential sources [1,2,7]. Bulk isotopic ratios of an oil, or refined product, may be affected by weathering. To minimize the problem, the isotopic composition of individual compounds can be determined by combined gas chromatography-isotope ratio mass spectrometry (GCIRMS/ GCIRMS has become a key and crucial aspect of environmental research (e.g. [1,8]) and is discussed in detail in a recent review Meier-Augenstein [9]. For spilled products consisting of a single component, GCIRMS is probably the only tool that can be used for attempted correlation purposes. Gasoline is a major groundwater contaminant in the United States and in 1996 there were 314,000 confirmed releases [10] of such substances from underground storage tanks, with many more unreported occurrences, and at approximately 40% of these sites the petroleum hydrocarbons reached groundwater [10]. GC and GCMS are of little use but GCIRMS provides a viable method for the discrimination of gasolines from different sources. MTBE, a gasoline additive, is a major groundwater contaminant in the US with 330,000 leaks involving MTBE from 1988 to 1998 [11]. The resistance of MTBE to degradation provides a potential indicator for the source of gasoline contamination. GC and GCMS cannot be used to distinguish sources of such a spill, but preliminary studies have indicated that GCIRMS can be used to discriminate MTBE from different sources based on its isotopic composition. The main purpose of this paper is to demonstrate how GCIRMS can be used to correlate weathered and unweathered samples from a variety of sources on the basis of the isotopic composition of the individual compounds.

3 Oil and Hydrocarbon Spills II: Modelling, Analysis and Control Experimental mehtods 2.1 Samples and weathering experiments Samples of weathered and unweathered crude oils, fuel oils, gasoline samples and tar balls from a variety of sources were used for this study. Weathering was simulated in the laboratory using a crude oil, aliquots of which were allowed to evaporate for varying periods of time. An aliquot of the oil was combined with distilled water and stirred for 2 months in order to simulate water washing. An active sewage sludge was added to the oil in a sand and water environment and aliquots of the biodegraded product were taken at regular intervals so that the main stages of biodegradation could be studied. 2.2 Sample preparation and analysis Most of the samples, except the refined products such as light fuel oils and gasolines, were characterized using the same protocol: (i)asphaltene precipitation by pentane; (ii)fractionationof the maltene fraction by HPLC; (iii) %-alkanes were isolated by urea adduction of the saturate fraction. 2.3 Gas Chromatography and gas chromatography-mass spectrometry Samples were analyzed by GC using a DB-1 fused silica capillary column (0.25mm i.d.; 1.0 urn film thickness) and temperature programmed from 40 to 300 C at 4 C/min. If necessary GCMS was performed with a Finnigan TSQ 70 combined with a Varian 3400 GC. 2.4 Gas Chromatography-Isotope Ratio Mass Spectrometry GCIRMS analyses were performed using a Varian 3400 gas chromatograph coupled to a Finnigan MAT 252 isotope ratio mass spectrometer via a combustion furnace, heated to 950 C, and a water trap. 2.5 Reproducibility and accuracy Accuracy and reproducibility of the isotopic values of individual compounds are mainly affected by the chromatographic resolution, column bleed, and complexity of the mixture. The accuracy of the data is monitored by the addition of fully deuterated %-alkanes Cg, C,o, C,6, C^, 34, 32 and C^, of known isotopic composition, to the samples being analyzed. The analysis of a saturate fraction gives an average reproducibility from 0.15 to 0.32%o. The results from this study are consistent with the accuracy and the reproducibility obtained in previous studies of similar samples [12,13].

4 214 Oil and Hydrocarbon Spills II: Modelling, Analysis and Control 3 Results and discussion Crude oils Oils from different geographical locations were characterized by GCIRMS to establish variations in the isotopic compositions of individual compounds in such samples. The discriminative nature of the isotopic compositions of individual n- alkanes in crude oils is illustrated in Figure 1. The averaged standard deviations calculated between these oils are higher than 0.55%o which is out of the range of the analytical error. The oils from Oklahoma and the Mahakam Delta have similar isotopic compositions which also illustrates the need to combine GCIRMS analyses with conventional analytical correlation techniques such as GC and GCMS wherever possible. n-alkanes mv)r~e\v-tmvir^c\vhff)*r)r^o,-4,-4?-4,-l<s<s(x(s(smmfn2r2 u u u u u u u u u o u u S S *o U Paris Basin Middle East Oklahoma Mahakam Figure 1. Isotopic composition of w-alkanes for five oils of different origins. The second fact that had to be established was that weathering would not cause changes in the isotopic composition of the individual compounds and compromise correlations between pollutants and suspected sources. Samples were subject to weathering by evaporation, waterwashing, and biodegradation and the artificially weathered samples were analyzed subsequently by GCIRMS and results compared to those from the initial oil. The initial oil showed a chromatogram largely dominated by therc-alkanesmaximizing around n-c^. The evaporation process removes %-alkanes with carbon numbers lower than 14 and the waterwashed oil (after 38 days) was depleted in %-alkanes, iso- and cyclo-alkanes with carbon numbers <15 consistent with literature results. Laboratory biodegradation results also were consistent with literature results, after four months w-alkanes have been removed and only a few iso-, cyclo-alkanes and some biomarkers can still be identified. The isotopic compositions of the individual %-alkanes of the four artificially weathered oils were measured and compared to those of the initial oil.

5 Oil and Hydrocarbon Spills II: Modelling, Analysis and Control 215 Despite the partial loss of /7-alkanes the results showed a good correlation between the evaporated, water-washed and unweathered oils. After 1 month, the isotopic composition of the biodegraded oil was slightly enriched in '"C compared to the initial oil and the standard deviation relative to the initial oil is slightly higher (0.4 l%o). To minimize interference from the unresolved hump (UCM) the saturate fraction of the biodegraded oils were fractionated into linear and branched/cyclic alkanes and the linear fractions analyzed by GCIRMS. The isotopic values of the linear fractions of the biodegraded and initial oils showed a very close correlation (ave.std. dev. 0.32%o). The lack of %-alkanes in the oil biodegraded for four months precluded GCIRMS analysis of the saturate fraction. The results of these experiments demonstrate that weathering does not affect the isotopic composition of individual components of an oil that are still resolvable in the chromatogram. Despite increased specificity, GCIRMS still suffers from a number of limitations in correlating moderate to severely biodegraded oils. GCIRMS is most effective when applied in the presence of /?-alkanes and a reduced background. In view of these limitations and that in the case of severely biodegraded oils the n- alkanes will have been removed, pyrolysis of the asphaltenes from heavily degraded oils provides an alternative method for correlation of biodegraded and non biodegraded oils. Off-line pyrolysis of asphaltenes isolated from the original and the biodegraded oils was performed as described elsewhere (14). The saturate + unsaturated fractions from the pyrolysates were subsequently analyzed by GCIRMS. The pyrograms of the oils biodegraded for 2 and 4 months were similar to that of the initial oil with a predominance of the light ends, maximizing at C^-,7. However, the two biodegraded samples show a higher contribution of n- alkanes around C%g and O,g whereas this contribution is very low in the asphaltenes of the initial oil. The correlation of these oils solely on the basis of the pyrograms of the asphaltenes would be extremely difficult. However, analyses of these fractions by GCIRMS showed the isotopic compositions of individual %-alkanes of these three fractions to be similar. These results confirm the ability of GCIRMS and the asphaltene pyrolysates to correlate severely biodegraded oils with their unweathered counterparts with an analytical error still acceptable despite the chromatographic problems encountered during the analyses of such complex mixtures. To investigate another major problem, oil from two sets of bird feathers coated with oil resulting from spills at sea have been examined. The correlation between the oil on the feathers and possible sources may be difficult since in many cases weathering will occur and lead to different fingerprints from the feathers vs. the suspected sources. As above, it has been shown by using GCIRMS it is possible to make the correlations and establish sources and the responsible parties. Extracts from oil covered bird feathers were analyzed by GCIRMS, and the data compared to the corresponding suspected sources. Correlations based on GC and GCMS were inconclusive. The first oil and bird feather extracts had the characteristics of a light fuel oil marked by the absence of biomarkers. The oil collected from the bird feathers was weathered and light hydrocarbons lost by evaporation and water-

6 216 Oil and Hydrocarbon Spills II: Modelling, Analysis and Control washing. The oils were analyzed by GCIRMS without any fractionation step prior to the analysis and their isotopic values were very similar. The standard deviation between the two samples (0.36%o) was close enough to the range of the reproducibility ( %o) to conclude that these two light fuel oils were derived from the same source. In the second example, the oils were more characteristic of a weathered crude oil with a predominance of long chain %-alkanes. The oil collected from the bird feathers contained lipids from the bird feathers, in addition to the hydrocarbons. The n-alkanes were isolated by urea adduction and results from the GCIRMS analyses of the n-alkanes showed a good correlation between the oil extract and the suspected source with a standard deviation of 0.23%o being within the range of reproducibility ( %o). These results provided a high level of confidence suggesting similar origins for the oils found on the bird feathers and the suspected sources. Spills of refined products including gasoline can cause major problems, particularly to groundwater or drinking water supplies. Much of the gasoline in the groundwater emanates from leaking underground storage tanks, or ruptured pipelines, but correlating spilled gasoline to its original source presents even greater problems than correlating crude oils. First many gasolines appear to be very similar to each other on the basis of their chromatographic behavior and effects of water washing and evaporation may be far more severe with gasolines than crude oils. Finally polar additives in gasolines complicate the situation since they are very water soluble and tend to become concentrated in the aqueous phase. a. A! b. jiu m. >Jvuu juu...ja I^AJL Figure 2. M/z 44 chromatograms showing the virtual identity of two different gasolines.

7 Oil and Hydrocarbon Spills II: Modelling, Analysis and Control 217 One additive, methyl tertiary butyl ether (MTBE), is responsible for the contamination of many municipal water supplies throughout the US as well as many lakes and rivers. Initially prescribed as an additive for gasolines to increase the oxygen content, MTBE is very recalcitrant in the environment and will remain in situ for many years. Correlation of groundwater MTBE with suspected sources is extremely difficult since MTBE is a single peak and GC and GCMS are virtually useless for such correlations. The only technique which appears to show some promise is GCIRMS and monitoring the isotopic composition of MTBE. MTBE can be isolated from water by purge and trap approach and the isotopic composition of the MTBE can be determined directly. As will be seen below there are some isotopic differences between MTBE from different suppliers. However the degree of variation in the isotopic composition of MTBE resulting from biodegradation remains to be evaluated. Chromatograms for two gasolines (DTX and EOK) are shown in Fig. 2 and are virtually identical. Sixteen compounds ubiquitous within the gasolines were selected and showed a wide range of variability with the greatest variability observed for benzene (peak number 76), 2,3-dimethylpentane (79), ethyl benzene (28), 1 -methyl-2-ethylbenzene (34) and 1,2,3-trimethylbenzene (36), all of which have std. dev. greater than 2.40 from the mean in agreement with previous results from the GCIRMS analyses of BTEX compounds from a large number of manufacturers(12,15). BTEX components in gasoline contaminated soils were analyzed by Kelley et al, 1997 (13) and it was reported that leaded and unleaded gasolines can be distinguished on the basis of the BTEX isotopic compositions. 8^C values for the BTEX components ranged from to %o (benzene), to %o (toluene), to %o (p & m- xylenes), and to (o-xylene). Four gasolines from various locations were analyzed and the results shown in Fig. 3. FOK and GOK are gas stations in close proximity to each other and from these results it can be suggested that they obtain their gasoline from the same supplier. Samples LEC and OEC are from different locations than FOK and GOK but again are probably from the same supplier. An attempt to survey the isotopic composition of "neat" MTBE was unsuccessful due to the manufacturers unwillingness to provide samples of their product. Two neat MTBE samples were obtained, and gave bulk carbon isotope numbers of-30.74%o and %o respectively. MTBE in a number of gasolines analyzed as part of this study gave reproducible carbon isotope data with a range of numbers from %o to %o. In the reproducibility study MTBE has a std. dev. of 0.57%o and MTBE with isotopic values outside of this range can be distinguished from each other. Whilst MTBE may not be used as a "silver bullet" it may be used in conjunction with classical techniques to differentiate sources of gasoline contamination.

8 218 Oil and Hydrocarbon Spills II: Modelling, Analysis and Control Peak Number Figure 3. Isotopic composition of individual components in the gasolines. Each point corresponds to a numbered component in Fig. 2. Three gasolines were allowed to weather via evaporation and water washing for varying periods of time. The composition of gasoline after 168h of weathering is clearly different from fresh gasoline would be impossible to use for spill to source correlations. At the end of one week only three compounds that were analyzed in the fresh gas are present in the evaporated residue (44, 45 and 46) of all three gasolines and these are all shown to be isotopically stable. These compounds are naphthalene, 1 -methyinaphthalene and 2-methylnaphthalene, all of which had a wide range of isotope numbers in the gasolines analyzed above (std. dev. 1.03, 1.16 and 1.42 respectively). Studies on the isotopic fractionation of low molecular weight hydrocarbons during microbial degradation have been conducted by Diegor et al [16] who found that whilst benzene has isotopic shifts of up to 2%o after degradation, toluene, ethylbenzene, naphthalene and flouranthene show no significant fractionation. However, research by Harrington et al. [15] has revealed that isotopic fractionation of aromatics (benzene, ethylbenzene and toluene) does occur during vaporization, with the vapor phase being enriched in ^C by up to 0.20 The isotopic stability of the naphthalenes after evaporation and biodegradation suggests that these are stable source markers for gasoline contamination. These preliminary results also suggest that the isotopic composition of MTBE can be used as a source marker, even after being solubilized into the water phase as would occur during contact with groundwater.

9 4 Conclusions Oil and Hydrocarbon Spills II: Modelling, Analysis and Control 219 Results from this study to evaluate the use of GIRMS as a potential monitoring and correlation tool for use in a variety of environmental problems have demonstrated that whilst GCIRMS should be used as a stand-alone technique, it raises the degree of confidence to which correlations are made to a higher level than those simply based on GC and GCMS data. This is particularly important when the samples being examined are weathered or heavily biodegraded, or single components. The successful use of GCIRMS depends on the premise that with weathering and biodegradation, the isotopic composition of individual components in a mixture will not be greatly affected. Hence as long as individual components remain in the weathered sample then isotopic compositions can be determined and used for correlation purposes. It also has the potential to be used for a wide range of samples and will be particularly useful for single components when conventional techniques are of little use. References [1] Macko, S.A.; Parker, P.L.; Botello, A.V. Environ. Poll. Bull, 2, , [2] Farran, A.; Grimalt, J.; Albaiges, J.; Botello, A.V.; Macko, S.E. Mar. Poll #%//., 18, , [3] Enganhouse, R.P.; Baedecker, M.J.; Cozzarelli, I.M.; Aiken, G.R.; Thorn, K.A.; Dorsey, T.F. /#,/. GeocAgm., 8, , [4] Lafargue, E; Barker, C. A.A.P.G. Bull., 72, , [5] Palmer, S.E. In Organic Geochemistry, eds M.H. Engel and S.E. Macko, Plenum Press, New-York, [6] Macko, S.A. and Parker, P.L. Mar.Environ. Research, 10, , [7] Hartman, B. and Hammond, D. Geochim. Cosmochim. Acta, 45, , [8] Bakel, A.J.; Ostrom, P.H.; Ostrom, N.E. Org. Geochem., , [9] Meier-Augenstein, W., J. Chromatogr. A842, , [10] EPA Audit Report, Consolidated Report on EPA's Leaking Underground Storage Program, # , [11] Koshland, C.P., R.F. Sawyer, D. Lucas, P. Franklin. Evaluation of Report to California State Legislature under S.B. 521: Effects of MTBE, [12] Dempster, H.S., L.B. Sherwood, and S. Feenstra, Environ. Sci. Technol., J7, , [13] Kelley, C.A., B.T. Hammer, R.B. Coffin, Environ. Sci. Technol., 31, , [14] Eglinton, T.I. Org. Geochem., 21, , [15] Harrington, R.R., S.R. Poulson, J.I. Drever,m P.J.S. Colberg, and E.F. Kelly, Org. Geochem., 30, , [16] Diegor, E.J.M., T. Abrajano, L. Stehmeier, T. Patel, L. Winsor, 19* International Meeting on Organic Geochemistry, Istanbul, Turkey, 1999, (extended abstract).