CEN methodology for oil spill identification

CEN methodology for oil spill identification

14

Paul G.M. Kienhuis*, Asger B. Hansen**, Liv-Guri Faksness†, Scott A. Stout‡, Gerhard Dahlmann§ *RWS-Laboratory, Rijkswaterstaat CIV, Lelystad, The Netherlands; **R&D – Organic Analysis, Haldor Topsøe A/S, Kongens Lyngby, Denmark; †Department of Environmental Technology, SINTEF Chemistry, Trondheim, Norway; ‡NewFields Environmental Forensics Practice, LLC, Rockland, MA, USA; §BSH-Laboratory-Sülldorf, Bundesamt für Seeschifffahrt und Hydrographie, Hamburg, Germany

14.1 Introduction Since 1991, the Nordtest method for oil spill identification (Nordtest, 1991)1 has formed an important forensic “platform” in relation to oil spill identification, not only in the Scandinavian countries, but also in other European countries following its recommendation in and adoption to the Bonn Agreement2 Counter Pollution Manual. The method basically incorporated a stepwise procedure including initial screening by GC/FID of all samples for characterization and to exclude obviously nonmatching candidate source samples. This step was followed by GC/MS fingerprinting of the selected samples (spill and potentially matching candidate source samples) recording a suite of key or target petroleum compounds. After comparing the GC/MS chromatograms in order to identify possible differences, a conclusion based on the chemical analysis would be reached as either identity (matching chromatograms) or nonidentity (nonmatching chromatograms). When evaluating the GC/FID and GC/MS chromatograms, weathering of the oil samples (i.e., changes in oil composition that take place after the spillage, including evaporation, dissolution, emulsification, photo-oxidation and biological decomposition) was taken into account as the criteria for identity excluding any differences arising from weathering. Ten to 15 years of experience with the Nordtest method, however, had shown some need for improvements. For example, the evaluation of chromatograms was generally achieved by qualitative means (i.e., visual comparison) and the influence of weathering was not always straightforward to interpret. These rendered the evaluation and final conclusions to be subjective, and hence questionable. In addition, over this time period advances in both analytical and interpretive methods had opened the possibility to obtain more quantitative, objective, and defensible means for verification of the results. Thus in 2000, Nordtest initiated a Revision of the Nordtest M ­ ethodology for 1

 ordtest is an institution under the Nordic Council of Ministers acting as a joint Nordic body in the field N of technical testing and standardization. 2 The Bonn Agreement is a multilateral agreement by North Sea coastal states, which together with the EU, will offer: (1) mutual assistance and co-operation in combating pollution, and (2) surveillance as an aid to detecting and combating pollution and to prevent violations of antipollution regulations. Standard Handbook Oil Spill Environmental Forensics. http://dx.doi.org/10.1016/B978-0-12-809659-8.00014-0 Copyright © 2016 Elsevier Inc. All rights reserved.

686

Standard Handbook Oil Spill Environmental Forensics

Oil Spill Identification with participation of the national forensic oil spill laboratories in Denmark, Finland, Norway, Sweden, and researchers in the United States. Phase 1 (2000–2001) of that project included: (1) a review and assessment of recently published literature on oil spill identification and petroleum geochemistry, and (2) an update of the existing 1991 Nordtest method into a technically more robust and legally defensible oil spill identification methodology through the introduction of laboratory techniques for sample preparation and cleanup, chromatographic analysis, and data evaluation tools. The resulting improved Nordtest methodology was still based on tiered GC/FID screening and GC/MS fingerprinting procedures, but had also included more objective criteria for sample comparisons by introducing quantitative diagnostic ratios between “key” petroleum compounds, that could be tested statistically, and adhered to stricter analytical protocols including QA/QC criteria (Faksness et al., 2002a). In Phase 2 of the project (2001–2002), the revised Nordtest methodology was widely evaluated through a round-robin exercise involving multiple laboratories (Faksness et al., 2002b). In 2002, Nordtest proposed that the revised methodology (Daling et al., 2002) be adopted as a new standard for oil spill identification by the European Committee for Standardization (CEN). In turn, CEN established a task force (CEN BT/TF 120) to evaluate the proposal and eventually prepare a new standard. The development of a new CEN standard was split into two work items: (1) sampling and (2) analytical methodology and interpretation of results, which required two working groups to produce two separate CEN Technical Reports on oil spill identification. In 2006, the two working groups produced a two-part guideline titled Oil Spill Identification – Waterborne Petroleum and Petroleum Products, viz., • •

Part 1 – Sampling – (CEN 15522-1, 2006a) Part 2 – Analytical methodology and interpretation of results (CEN 15522-2, 2006b)

In 2005 the Bonn Agreement2 requested the laboratories of countries around the North Sea to improve their cooperation, promulgated by problems that occurred during the Tricolor incident (2002) in the British Canal. Each country participating in the Bonn Agreement assigned a laboratory to be responsible for the oil spill cases in their area, and at a meeting of national lab representatives in Ostend (Be), the BonnOil Spill Identification Network of experts (Bonn-OSINet, 2005) was formed (Dahlmann and Kienhuis, 2015). Bonn-OSINet was initiated with the objective to improve the quality of the laboratories and to stimulate cooperation and mutual assistance. All members of the previous CEN BT/TF 120 task force also participated in BonnOSINet. In 2010, it was decided to publish a revised version of the 2006 CEN guidelines. Part 1 was reviewed and found to not require updating, while Part 2 was further refined and republished in 2012 (CEN/Tr 15522-2, 2012). In 2005, SINTEF (Norway) produced an oil mixture to facilitate the identification of target compounds and compound groups in oil samples. The mixture is a combination of three crude oils (from Russia, Sicily, and the North Sea) and a heavy bunker oil (IFO-180). The mixture can be used as a reference oil because it contains all of the compounds mentioned in the CEN guideline (and this chapter). The oil mixture can be obtained from SINTEF together with all relevant and integrated chromatograms, which can facilitate implementation of the CEN method at laboratories performing

CEN methodology for oil spill identification

687

oil spill identification. The chromatograms shown in figures throughout this chapter (and in the CEN/Tr 15522-2 [2012] guideline) are based on the SINTEF reference oil. In 2012, BSH (Germany) introduced the web-based COSI database to the OSINet participants. This method is more thoroughly described in Chapter 15. Briefly, after uploading complete FID and GC/MS chromatograms of samples in csv or cdf format, the database can be used to compare and evaluate samples according to the CEN 2012 method. At present, the COSI database contains more than 2600 samples including 300 crude oils. The database offers the possibility to search for best matching samples within the database. This search option, based on the combined ratios of each sample, offers the possibility to receive an impression of the “uniqueness” of a spill sample (see Dahlmann and Kienhuis, 2015 and Chapter 15). The present chapter is focused on presenting the technical highlights of the 2012 CEN Guideline – Part 2. We will describe the tiered analytical approach, the use of percent weathering (PW) plots, the comparison of diagnostic target compound ratios and the data treatment. Sampling techniques and handling of oil samples prior to their arrival at the forensic oil spill laboratory, as described in Part 1 (CEN/TR 155221: 2006a), will not be covered here, but are largely summarized in Chapter 2.

14.2 Intercalibrations The CEN methodology (both 2006 and 2012 versions) has been evaluated since 2002 via annual intercalibration exercises (Round Robins, or RRs) involving multiple national laboratories and other researchers around the world. The RRs have been an important tool to test the methodology and laboratories, but has also provided opportunities to learn from each other and to improve knowledge regarding protocols, weathering, and different petroleum compositions (crude oils and products), as described in Table 14.1. In Phase 2 of the Revision of the Nordtest Methodology project (2001–2002), the updated methodology was evaluated through a RR exercise arranged by SINTEF. In 2004 and 2005, RWS-RIZA and BSH organized two RR exercises. Since 2006, organizing an annual RR has become a core part of the activities of Bonn-OSINet. The results of each exercise are discussed at an annual meeting, together with many other issues relevant for oil spill identification. The lessons learned have been taken into account for refining and updating the CEN methodology. See Bonn-OSINet web page for a free download of the summary reports of the past RR exercises. (Note: This chapter has several references to the RR summary reports; for example, a reference to the summary report of Round Robin 2011 will be indicated by [Bonn-OSINet; RR2011]).

14.3  Objective and scope of the CEN methodology The objective of the CEN methodology is to provide a forensic tool for the identification of waterborne oil by comparing samples from spills with those of suspected sources. The methodology should be capable of providing both administrative and

688

Standard Handbook Oil Spill Environmental Forensics

Table 14.1  Round Robins organized to evaluate the CEN methodology Year

Product

Subject

Labs

2002 2004

Crude – HFO (NO) Gas oil case (NL)

12 16

2005 2006

Bilge (NL) Russian Crude (DE)

2007

HFO (DE; NL)

2008 2009 2010 2011

Crude (NO) Bilge (ES) HFO – Crude (CN) HFO (Fr)

2012

Crude (UK)

2013

HFO – lubricating oil (SE) Crude oil (FR)

Testing the proposed updated Nordtest method Product type study. Added to the draft CEN method Evaporation and ratio calculations Recognition of the oil type Comparability of data Tricolor. PW plots, integration (area-height), variability PW plots for the estimation of weathering Mixing and biological degradation Evaporation and biological degradation Prestige and Erica spills. Heavy weathering 10-year-old samples. IMOF method tested Complexity of Nigerian oils. Photo-oxidation and dispersants Real case from Sweden. Weathering, mixing, and background Dissolution and burning

2014

12 13 19 23 24 24 25 26 27 31

Samples have been provided by the indicated countries (between brackets).

legal support to the prosecution of an offender (“potential responsible party,” or PRP) that has violated national or international regulations by illegally and/or accidentally discharging mineral oil into the environment. The scope is to provide a methodology to identify waterborne oils spilled in marine, estuarine, and aquatic environments based on detailed analytical and processing procedures for the comparison of samples from spills with those of suspected sources. When suspected sources are not available, the methodology may still be used to characterize the spilled oil, and determine oil type and likely origin, which may help identify a PRP. The methodology is restricted to petroleum and petroleum products containing a significant proportion of petroleum hydrocarbons with boiling points above 200°C (n-C11+) such as the following: • • • • •

Crude oils Light refined products like diesel oils or gas oils Heavy refined products like heavy fuel oils, bunker oils, and vacuum residues Lubricating oils Mixtures of bilge and sludge samples

Still, while the general concepts of the methodology have a limited applicability for some kerosenes and condensates, it may not be applicable for gasoline due to its volatile nature (see Chapter 11).

CEN methodology for oil spill identification

689

The CEN methodology is not intended for oil spilled to groundwater and soil. This is because the chromatograms of oil extracted from soil and groundwater may contain reduced and/or additional compounds compared to the candidate source sample. Evaluating such samples in the methodology would require additional extraction and cleanup methods together with an accounting of which compounds may have been reduced and/or added due to matrix effects, either of which may affect the comparison and final conclusion. Therefore, such issues are beyond the scope of the existing CEN methodology described herein. However, in cases where oil is spilled to groundwater, soil, or sediment – and appropriate sample preparation and the potential for compound reductions and/or additions are appreciated – a positive match achieved using the CEN methodology would be valid.

14.4  Strategy for identifying the source of an oil spill When an oil spill has been observed, samples should be collected from the spill and from potential responsible parties, such as suspected ships or other sources, in due course by appropriate authorized personnel (Chapter 2). All samples should subsequently be sent by either an authorized “sampling coordinator” or directly to the analytical laboratory for forensic characterization and potentially for identification of the spill’s source. In the context of this CEN methodology, the process of identifying the source of a spilled oil implies a comparison be made between the chemical compositions of the spilled oil and all candidate source samples. Conceptually, two results can be achieved in forensic oil spill investigation – match and nonmatch, depending on whether spill and candidate source samples are identical or nonidentical. Match, in theory, requires all measurable data to match “exactly.” For a laboratory, however, it is practically and technically impossible to measure all physical properties and to compare every compound (group) in each sample (Albaigés et al., 2015a). However, GC/FID and GC low-resolution MS are cost-effective techniques available nowadays in almost every analytical laboratory. GC/FID and/ or GC/MS operated in full scan mode yield chromatograms that provide an overall impression of a sample, while GC/MS operated in the selected ion monitoring (SIM) mode can be used to analyze a large number of individual compounds or compound groups with a good specification, sensitivity, and robustness. Other analytical tools such as GC × GC/FID (e.g., Aeppli et al. 2014; Radovic´ et al. 2014; see also Chapter 8) and GC/MS–MS (Shang et al., 2014) have been considered to be added to the method, or even to replace GC/FID and GC/MS. However, until these other tools become more widely used and proven, there is no plan to incorporate these as essential to the ­methodology. In practice, two samples are considered to be a positive match if no statistically significant differences (at the 95% confidence level) among a large number of diagnostic metrics determined by GC/FID and GC low-resolution MS analyses are present, which cannot be explained by weathering, mixing, or heterogeneity. This approach, that is, looking for differences in diagnostic signatures instead of similarity among ­every

690

Standard Handbook Oil Spill Environmental Forensics

possible signature, is conceptually more logical and is more practically and technically achievable. As such, only distinct differences between samples can be proved. Therefore, when no statistically significant differences between samples are observed a “positive match to a high degree of scientific certainty” should be concluded. To be able to distinguish between nonmatching but closely related samples, it is essential that the analytical systems have a low variance for the samples and metrics involved. The methodology requests a range of 85–118% for the data points of a PW plot and a maximum RSD of 5% for the ratios. To demonstrate that this is valid, at least two and/or 10% of the samples (extracts) of a spill case have to be analyzed in duplicate. The duplicates must show a clear positive match with a low variance, before the spill samples can be compared. If these performance criteria cannot be met, the integration and analytical system must be checked and the samples reanalyzed. In Europe, forensic oil spill identification is performed by laboratories that analyze oil samples on a regular basis, and also by laboratories that only compare samples a few times a year. In the past, the common practice had been to analyze oil samples qualitatively and compare the chromatograms and ion fragmentograms visually – as per the original Nordtest (1991) protocol. As noted previously, the results of such comparisons depend heavily on the experience of the laboratory personnel and data interpreters, which at laboratories that only rarely analyzed oil samples, was not optimal. To aid in this regard, the CEN method introduced the use of percentage weathering plots (PW plots) and diagnostic ratios3 as additional and more objective and defensible tools for comparison. The criteria for selecting compounds (or compound groups) and ratios take into account their behavior upon weathering and their variability among oils of different types (e.g., fuel types) and petrogenic origin (North Sea versus Middle East). To reduce the analytical variance, ratios are preferably generated by using the peak area of compound groups and the peak height of single compounds, the latter of which are preferably recorded by the same m/z value and approximately within the same retention time window. The resulting ratios are successively compared using the repeatability limit (r95%) as a test method. Optionally, diagnostic ratios can also be generated on the basis of chromatographic peaks that have been fully (absolutely) quantified using external and/or internal standards. The use of PW plots and diagnostic ratios for comparison of oil samples is based on GC/FID data of n-alkanes and isoprenoids and on GC/MS data of a suite of alkylated polycyclic aromatic compounds (PACs) and petroleum biomarkers. Collectively, these targeted compounds provide a range of markers capable of monitoring different types of weathering and revealing “genetic” differences between oil from different origins and between different product (fuel) types. The method prescribes a product-specific minimum list of “normative” compounds and ratios that have to be analyzed and used for comparison. At the same time, the method allows for flexibility in the use of additional compounds and ratios that are relevant for the samples involved in a particular oil spill case. 3

 iagnostic ratios (DR) – ratios between the peak height of single compounds or the area of groups of D compounds selected for their diversity in the chemical composition in petroleum and petroleum products and their reported response to weathering and degradation processes.

CEN methodology for oil spill identification

691

Before integrating the target compounds or compound groups required for generating the selected diagnostic ratios, a visual inspection of the relevant ion fragmentograms should be carried out in order to eliminate those target compound peaks not present in a sufficient amount to fulfill the required signal-to-noise (S/N > 3–5) criterion, and hence not capable of generating robust diagnostic ratios. A visual comparison of the ion fragmentograms of m/z 191, 192, 217, 218, 231, and 216 is also recommended to exclude obviously different samples and recognize any highly diagnostic compounds specific for the actual case (e.g., unusual biomarkers) that may not be prescribed target analytes. After the comparison and evaluation of the PW plots and the diagnostic ratios, a second visual comparison of the ion fragmentograms of the relevant samples one by one, should also be carried out to verify (ground truth) any conclusion reached.

14.5  Visual characterization and preparation/cleanup of oil samples When arriving at the analytical laboratory, all samples should be carefully examined and described with respect to type (e.g., spill sample or source/reference sample), matrix (e.g., water, sand, feather etc.), amount, container (glass, plastic, etc.) and general condition. Eventually, all samples should be photographed to assist the visual description and document their condition upon arrival at the laboratory. Any sign of sample compromise or “missing links” in the chain-of-custody should be reported to the sampler/sample coordinator immediately for resolution. The applied procedures for forensic identification of spilled oil must strictly observe that any manipulation performed during sample preparation and cleanup can alter its chemical composition, and thus potentially weaken the results as evidence. Therefore, sample preparation and cleanup should generally be kept to a minimum, and in any case precisely described in any final report. After removal of any pieces of debris (e.g., wood, fabric, feathers etc.), preparation is generally performed by diluting the sample in solvent to appropriate concentrations for GC/FID and GC/MS analysis. For samples contaminated with polar compounds (e.g., bird feathers), soot particles (used lubrication oil), or include a high amount of heavy components (e.g., asphaltenes in heavy bunker oil) a simple cleanup on a silica/alumina or magnesium silicate (Florisil®) column must be applied. Descriptions of preparation and cleanup of oil samples are beyond the scope of this chapter, but more detailed procedures for the different sample types can be found in the CEN 15522-2: 2012 (CEN 2012) guideline.

14.6  Decision chart for identifying the source of spilled oil The methodology for identification of oil spills is divided into two levels of analytical procedures and data treatments according to the decision or flow chart shown in Figure 14.1.

692

Standard Handbook Oil Spill Environmental Forensics

Figure 14.1  Decision/flow chart of the CEN oil spill identification methodology.

CEN methodology for oil spill identification

693

The flow chart guides the user through the individual steps by linking each step of the procedure with the operation performed (analysis/evaluation) at the previous step and its resulting decision until a final conclusion regarding identity can be made. In the flow chart, squared boxes refer to operations to be performed, and hexagon boxes refer to evaluations/conclusions to be made. The flow chart is comprised of the following steps: GC/FID screening of all samples – visual characterization/classification of oil types – GC– PW plot to evaluate variance and weathering – calculation of acyclic isoprenoid diagnostic ratios • GC/MS fingerprinting of selected samples (spill(s) and candidate source(s) – visual evaluation of chromatograms – MS–PW plot to evaluate variance and weathering and possibly mixing – comparison of diagnostic ratios (using the repeatability limit criteria) • Conclusion and reporting •

14.7  Level 1 – GC/FID screening After sample preparation, all spill and suspected source sample extracts are initially characterized by GC/FID screening on a nonpolar column. When a GC/FID is not available, a GC/MS operated in full scan mode (m/z 50–500) can be used. This analysis is intended to give information about the injection concentration and a general impression of the type of sample in a short time. A column with an internal diameter of 0.18 mm with hydrogen as a carrier gas will make run-times of only 15 min possible, but it can be easily overloaded. An internal diameter of 0.25–0.32 mm and a length of 15–30 m is more robust and will give run-times of 45 min and achieve better resolution between some close-eluting peaks. For example, the method has a minimum requested value of 0.7 for the peak pair n-C17 – pristane (Figure 14.2). It is advised to inject samples at a mid-level concentration in such a way that the compounds with the highest concentrations are not overloaded, while compounds with a low concentration can still be detected. CEN 2012 gives guidance in the injection concentrations of petroleum products, but for spill samples taken with, for example, an EFTE net, the injection concentration is unknown. If the GC/FID analyses reveal that the injection concentration is not at the appropriate level, it is strongly advised to adjust the concentration and to repeat the analyses in order to realize a low variance in the PW plots and the ratio comparisons for GC/FID and GC/MS.

14.7.1  GC/FID – level 1.1: visual inspection and elimination GC/FID chromatograms provide a descriptive “picture” of the dominating petroleum hydrocarbons in the oil sample, for example, the overall boiling range and prominence of individual resolved n-alkanes and major isoprenoids (Figure 14.2). GC/FID chromatograms also provide information on the weathering extent of the spilled oil, and on any “characteristic features” or “contaminating” components present in the samples. For example, biodiesel containing fatty acid methyl esters

694

Standard Handbook Oil Spill Environmental Forensics

Figure 14.2  GC/FID chromatogram of a mildly weathered fuel oil (diesel) displaying the dominating n-alkanes together with the acyclic isoprenoids, pristane, and phytane. A weathering test by means of a bar chart (A) and a GC–PW plot (B) are inserted.

(­ FAMEs) can be easily observed via GC/FID when it is mixed with petroleum diesel, with concentrations up to 5% being very common (Fuller et al., 2013). Figure 14.2 shows some extra peaks in between some n-alkanes, with the double peak just before n-C21 acting as a good marker for the presence of biodiesel (DeMello et al., 2007). Generally, the n-alkanes are the most characteristic and dominating peaks distributed regularly over the entire retention interval, except for extensively weathered oil samples such as water samples collected from thin oil films (sheens), highly biodegraded crude oils (due to biodegradation in the reservoir prior to spillage), and certain refined petroleum products (like lubricating and hydraulic oils).

14.7.2  GC/FID – level 1.2: evaluation of weathering If the chromatograms of a spill and candidate source oil appear different and the differences could possibly be caused by weathering, a “weathering check” is performed. This can be achieved by integrating (by peak height or area) the n-alkanes and isoprenoids (pristane and phytane) in the GC/FID chromatograms, then normalizing the peaks to nonweathered compounds (e.g., the mean of n-C20 to n-C24), and displaying the normalized peaks in a bar chart (e.g., prepared by Excel™) for comparison of the spill sample with a nonweathered source sample. An example of this type of plot is

CEN methodology for oil spill identification

695

shown in the inset to Figure 14.2A. Such plots reveal whether each lower n-alkane is more evaporated due to its lower boiling point. The percentage of each compound in the weathered sample relative to the same compound in the nonweathered sample is then calculated and these percentages are plotted as a so-called “percentage weathering plot” (GC–PW plot), as shown in the inset to Figure 14.2B. The sinuous shape of the curve on the GC–PW plot for these two samples confirms that evaporation has caused the reduction of the low boiling compounds up to n-C14. At the same time, the plot shows that the rest of the n-alkanes and isoprenoids plot along the 100% line, indicating a low variance and high similarity between the surface water and fuel tank samples. Notably, the bar-chart presentation and the GC–PW plot of the n-alkane distribution in oil samples may also provide information on possible wax/paraffin redistribution as a part of the weathering process (Strøm-Kristiansen et al., 1997). Alternatively, although less informative than the GC–PW plots, Figure 14.3 gives another example of a qualitative “weathering check” that can be achieved simply by overlaying chromatograms of a spill sample and a suspected source (in this case, heavy fuel oil Bonn-OSINet; RR2007). The normalization/manipulation of the chromatograms to a comparable attenuation can easily be done by expanding one of the chromatograms vertically until the “baseline” and the peaks in the n-C20 to n-C24 range are at the same level.

Figure 14.3  GC/FID chromatogram overlay of a weathered heavy fuel oil from the Tricolor case (2002).

696

Standard Handbook Oil Spill Environmental Forensics

If the comparison of the GC/FID chromatograms of the spill samples with the candidate source samples reveals differences in the n-alkane distribution, the unresolved complex mixture (UCM) shape, and/or in the acyclic isoprenoid ratios that obviously are not attributable to weathering, and that are significantly higher than the analytical variance, then such source samples are concluded to be nonmatching (i.e., “nonidentity” or nonmatch is achieved) and can be ruled out and eliminated from additional levels of analysis (Figure 14.1). If, however, there are any doubts about the conclusions, the samples should still be considered as potential sources and analyzed further in accordance with Level 2 of the flow chart.

14.7.3  GC/FID – Level 1.2: diagnostic ratios GC/FID chromatograms also enable the calculation of some diagnostic ratios such as n-C17/pristane, n-C18/phytane, and pristane/phytane ratios, unless the spill sample is too weathered, or the amount of these compounds is low relative to the UCM hump, as can often be observed in bunker oils. These ratios can also be indicative of biodegradation of the spilled oil, as they can monitor the effect of microbial degradation at the spill site by the preferential loss of n-alkanes relative to isoprenoids.

14.8  Level 2 – GC/MS fingerprinting At this level, selected spill samples and those candidate source samples that have not been eliminated by the GC/FID screening (Level 1) are analyzed by GC/MS operated in the selected-ion-monitoring mode (GC/MS-SIM). The GC/MS analysis at this level is used for characterizing and assessing the content and distributions of a suite of diagnostic and target alkylated PAC and petroleum biomarker analytes, from which the MS–PW plots and recommended diagnostic ratios can successively be generated. After injection concentration adjustment based on the GC/FID results (as needed), the selected samples are analyzed on a slightly polar capillary column to improve the separation of the PACs. A column with an internal diameter of 0.18 mm (L = 15 m) up to 0.32 mm (L = 60 m) and a stationary phase of 5% phenyl 95% dimethylpolysiloxane (e.g., DB-5; HP-5ms and HP-5) is advised. A phase of 5% phenyl 95% dimethyl arylene siloxane (e.g., DB-5ms) is not able to separate the close-eluting pairs n-C17-pristane and n-C18-phytane properly (Bonn-OSINet; RR2007). CEN 2012 Annex B gives a detailed description of the SIM m/z sections in which the principle of the Kováts index is used to indicate the different sections. A temperature program with a gradient of 4–6°C/min is advised, which facilitates compounds eluting up to gammacerane to elute before the final isothermal section starts (Bonn-OSINet; RR2007). A minimum resolution of 0.7 is requested for elution pair 3-MPhe–2-MPhe (­Figure 14.4) and of 0.9 for the elution pair 31abS–31abR (Figure 14.5). Chromatograms intended for uploading to COSIweb (see Chapter 15) need to be “identical” in separation and retention times. Samples have to be analyzed with a DB-5 column with a length of 30 m and an ID of 0.25 mm. In order to obtain retention times of 30.00 ± 0.02 min for 3-MPhe and 47.80 ± 0.02 min for

CEN methodology for oil spill identification

697

Figure 14.4  GC/MS ion fragmentograms recorded in the SINTEF standard oil mixture. (A) Methylphenanthrenes (m/z 192); (B) methyldibenzothiophenes (m/z 198); (C) C2phenanthrenes (m/z 206); (D) C2-dibenzothiophenes (m/z 212); (E) C3-phenanthrenes (m/z 220); (F) C3-dibenzothiophenes (m/z 226); (G) C4-phenanthrenes including retene (m/z 234 and 219); (H) methylfluoranthenes/methylpyrenes/benzofluorenes (m/z 216); (I) C3-chrysenes (m/z 270).

698

Standard Handbook Oil Spill Environmental Forensics

Figure 14.5  GC/MS ion fragmentograms (A and B) of pentacyclic triterpanes (hopanes) and tricyclic triterpanes recorded at m/z 191; (C and D) of tetracyclic rearranged (diasteranes) and regular 17a(H)-steranes, and 17b(H)-steranes recorded at m/z 217 and m/z 218, respectively; (E) of triaromatic steroids recorded at m/z 231; and (F) of sesquiterpanes recorded at m/z 123 in the SINTEF standard oil mixture.

CEN methodology for oil spill identification

699

30ab hopane, the retention times of the last standard of each sequence must be checked. CEN 2012 Annex B gives detailed information on how to adjust the initial time and slope of the temperature program. In general, only the initial time has to be adjusted after a few sequences to keep the retention times required for COSIweb comparisons. To test whether the GC/MS chromatograms can also be used for the IMOF method (a multivariate method to compare GC/MS chromatograms; see Chapter 16) over a longer period of time, chromatograms of a standard prepared from Brent crude oil, analyzed before and after each sequence, have been collected over a time period of 2 years by RWS-lab (NL). IMOF (Christensen et al. 2004; Chapter 17; Bonn-OSINet, RR2011) showed that a retention time adjustment was barely needed to have an exact overlay of the peaks of all compounds, resulting in the possibility to compare chromatograms with IMOF from different sequences. But even when automatic chromatogram evaluation with COSIweb or IMOF is not intended, it is strongly advised to work with fixed retention times. Checking and adjustment after each sequence takes only a few minutes, and because compounds always elute at the same retention time, allows for better recognition and consistent integration of target compounds (or compound groups).

14.8.1  GC/MS – level 2.1: visual inspection and elimination After analyzing the selected samples by GC/MS, the recorded chromatograms ­(Figures 14.4 and 14.5) of at least m/z 85 (alkanes), 191 (triterpanes and hopanes), 192 (methyl-phenantrenes), 216 (methyl-fluoranthenes/pyrenes/benzofluorenes), 217 and 218 (diasteranes and steranes), and 231 (triaromatic steranes) should be visually checked for characteristic features or obvious differences, which could possibly eliminate any suspected source sample from the candidate sources. If there is any doubt whether observed differences are due to genuine differences or to the effect of weathering and degradation, the samples should still be considered as potential sources and integrated in accordance with Level 2.2 of the CEN method’s flow chart (Figure 14.1).

14.8.2  GC/MS – level 2.2: peak measurements Analytical data from PAC and biomarker compounds form the basis for generating MS–PW plots and ratios. It is generally recommended that applied diagnostic and normative (Section 14.8.2.3) ratios are based on single compounds recorded at the same m/z value (e.g., m/z 192: NR_2-MPhe/1-MPhe), to eliminate the mass spectrometer’s varying response for different ions. In many cases, however, it may be of diagnostic value also to include ratios based on different m/z groups (e.g., m/z 212/206: DR_C2Dbt/C2-Phe) to assess different levels of various compound groups. For the PACs, diagnostic ratios based on peak heights are generally recommended as most compounds are well-resolved peaks with smooth baselines and can be easily integrated for height. However, some diagnostic ratios are based on whole isomer groups (e.g., DR_C2-Dbt/C2-Phe), which require integration of the appropriate compound groups based on the total area. For the biomarkers, integration of peak heights are generally recommended; some diagnostic ratios are based on peaks that

700

Standard Handbook Oil Spill Environmental Forensics

may not be well-resolved and/or have noisy baselines, in which cases, ratios based on peak heights are more robust. When integrating individual peaks, the use of coeluting peaks should be avoided, if possible. For example, if the peaks of C29Ts and C29ab coelute (Figure 14.5), use the peak height of the combined peaks. In practice, diagnostic ratios can be calculated either as A/B or A/(A + B), where the latter will always give values in the range of 0–1. However, the use of ratios compared to single components generally lowers the analytical variance, and ratios calculated as A/(A + B) result in lower relative standard deviations (RSDs) compared to ratios calculated as A/B. On the other hand, ratios calculated as A/B give constant RSDs independent of the numerical values of A and B, which is not the case for A/(A + B) ratios wherein the RSDs depend on the actual ratio value (Faksness et al. 2005; BonnOSINet; RR2004; Sect. 3.3). Thus, ratios calculated as A/(A + B) are less suitable for a comparison based on repeatability because the confidence interval has to be adjusted to reflect that the RSD depends on the actual value. Therefore, use of the repeatability limit within the CEN 2012 methodology requires that diagnostic ratios (DR) between two distinct and well-resolved peaks (A and B) should be based on the ratio formula: • •

DR = A/B or DR = 100 × A/B (%; depending on the preference of the user)

14.8.2.1  Diagnostic ratios derived from alkylated polycyclic aromatic compounds Petroleum originating from different oil fields and petrogenic provinces generally has sufficient differences in their distribution of alkylated PAC isomers to be of diagnostic value. Several PAC diagnostic ratios have been described and applied for oil spill identification (Wang et al., 1999, 2008; Weiss et al., 2000; Daling et al., 2002; Stout et al., 2002; Stout and Wang, 2008; Wang and Fingas, 2003). The most commonly used diagnostic groups of alkylated PAC isomers are the phenanthrenes (3-ring) and the heterocyclic dibenzothiophenes (3-ring) that can be found in most sample types (c.f., Section 14.3 and Figure 14.4). However, other alkylated PAC isomer groups can be of diagnostic value including the fluorenes (3-ring) and the chrysenes (4-ring). The CEN methodology also utilizes ratios of the methylfluoranthenes/methylpyrenes/benzofluorenes (4-ring) because compounds within this group show significant variation between different oils. Thus, in total, 12 diagnostic ratios derived from alkylated PAC isomers are recommended, as listed in Table 14.2. To facilitate and ensure proper identification of the chromatographic peaks used for generating the recommended diagnostic ratios, the CEN methodology includes a series of relevant ion-chromatograms as shown in Figure 14.4A–I. Of particular importance are the methyl-phenanthrenes (m/z 192), as they may be used to differentiate between crude oils and heavy bunker fuel oils (Figure 14.4A). For example, it has been reported that in crude oils, the first pair of peaks (i.e., the 3-methyl- and 2-methylphenanthrenes) are generally less abundant then the second pair of peaks (i.e., the 9-/4- and 1-methylphenanthrenes), while in heavy fuel oil (HFO) treated by a catalytic cracker at the refinery or mixed with products of a catalytic cracker, this is reversed (Dahlmann, 2003). This can be confirmed by the fact that retene (Figure 14.4G m/z 234 and m/z 219, for confirmation) is not present in this type of HFO.

CEN methodology for oil spill identification

701

Table 14.2  Normative and recommended diagnostic ratios (DR) derived from alkylated PACs* Ratio name

Definition

m/z value

NR_2-MPhe/1-MPhe** NR_4-MDbt/1-MDbt

2-Methylphenanthrene/1-Methylphenanthrene 4-Methyldibenzothiophene/1-Methyldibenzothiophene C2-Dibenzothiophenes/C2-Phenanthrenes C3-Dibenzothiophenes/C3-Phenanthrenes C3-Dibenzothiophenes/C3-Chrysenes Retene (7-isopropyl-1-methylphenanthrene)/ Tetra-Methyl-phenanthrene Benzo[b]naptho[1,2-d]thiophene/Tetra-methylphenanthrene 2-Methylfluoranthene/4-Methylpyrene Benzo[a]fluorene/4-Methylpyrene Benzo[b + c]fluorene/4-Methylpyrene 2-Methylpyrene/4-Methylpyrene 1-Methylpyrene/4-Methylpyrene

192 198

DR_C2-Dbt/C2-Phe DR_C3-Dbt/C3-Phe DR_C3-Dbt/C3-Chr NR_Retene/T-M-phe NR_BNT/T-M-phe NR_2-MFL/4-MPy NR_BaF/4-MPy NR_B(b + c)F/4-MPy NR_2-MPy/4-MPy NR_1-MPy/4-MPy

212/206 226/220 226/270 234 234 216 216 216 216 216

* C# here denotes the number of carbon atoms in the PAC alkyl substituent, that is, C2 could either denote dimethyl or ethyl and C3 either trimethyl or ethyl-methyl. ** In cases where the double peaks of the methylphenanthrenes are not properly resolved, this diagnostic ratio could alternatively be generated from area of the double peaks, for example, DR_(3-MPhe + 2-MPhe)/(9/4-MPhe + 1-MPhe).

The isomer compound cluster recorded by m/z 216 (Figure 14.4H) may also provide valuable diagnostic characteristics, as these have been shown to be relatively resistant to biodegradation and suitable especially for comparing light fuel oil samples like gas oils. More detailed features of individual oil types encountered in most environmental oil spill situations have been characterized and described by Dahlmann (2003) and are described more fully in CEN 2012 Annex H.

14.8.2.2  Diagnostic ratios derived from petroleum biomarkers Biomarkers are naturally occurring, ubiquitous, and stable hydrocarbons that are present in crude oils and most petroleum products. They are complex “molecular fossils” derived from once-living organisms. Biomarkers’ specificity, diversity, complexity, and relative high resistance to weathering make them extremely useful as diagnostic “markers” in the characterization and differentiation of spilled oils and candidate source oils (Stout et al., 2000; Stout and Wang, 2008; see also Chapter 3 and 4). The most common biomarkers used by organic geochemists include sesquiterpanes (e.g., drimanes), triterpanes (e.g., hopanes), diasteranes/steranes (e.g., diacholestanes/cholestanes) and mono- and triaromatic steroids (Peters et al., 2005; Yang et al., 2013). By exploiting the experience gained by petroleum exploration and production geochemistry, combined with the results of an extensive analysis of a large number of oils (Faksness et al., 2002a), a suite of diagnostic biomarker ratios have been selected as technically defensible indices useful in differentiating among qualitatively similar oils and comparing spilled oil to the available candidate source oils.

702

Standard Handbook Oil Spill Environmental Forensics

Table 14.3  Normative (NR) and recommended diagnostic ratios (DR) derived from tri- and pentacyclic triterpanes Abbreviation

Compound name

m/z value

C23Tr C24Tr C25Tr C28 (22S) C28 (22R) C29 (22S) C29 (22R) 27-Ts 27-Tm 28ab 25norC30ab 29ab 29Ts 30d 29ba 30O 30ab 30ba 31abS 31abR 30G 32abS 32abR 33abS 33abR

C23 tricyclic diterpane C24 tricyclic diterpane C25 tricyclic diterpane (a + b) C28 tricyclic triterpane (Cheilanthane) C28 tricyclic triterpane (Cheilanthane) C29 tricyclic triterpane (Cheilanthane) C29 tricyclic triterpane (Cheilanthane) C27 18a(H)-22,29,30-trisnorneohopane C27 17a(H)-22,29,30-trisnorhopane C28 17a(H),21b(H)-28,30-bisnorhopane C29 17a(H),21b(H)-25-norhopane C29 17a(H),21b(H)-30-norhopane C29 18a(H)-30-norneohopane C30 15a-methyl-17a(H)-27-norhopane (diahopane) C29 17b(H),21a(H)-30-norhopane (normoretane) C30 18a(H)-oleanane C30 17a(H),21b(H)- hopane C30 17b(H),21a-(H)-hopane (moretane) C31 17a(H),21b(H),22S-homohopane C31 17a(H),21b(H),22R-homohopane C30 Gammacerane C32 17a(H),21b(H),22S- bishomohopane C32 17a(H),21b(H),22R-bishomohopane C33 17a(H),21b(H),22S- trishomohopane C33 17a(H),21b(H),22R-trishomohopane

191 191 191 191 191 191 191 191 191 191 191 191 191 191 191 191 191 191 191 191 191 191 191 191 191

Ratio name

Definition

Ratio name

Definition

DR_C23Tr DR_C24Tr DR_C25Tr DR_C28 DR_C29 NR_C27-Ts NR_C27-Tm DR_Ts/Tm

C23Tr 30ab C24Tr/30ab C25Tr/30ab C28(22R)/30ab C29(22R)/30ab C27-Ts/30ab C27-Tm/30ab C27-Ts/C27-Tm

NR_C28ab DR_25norC30ab NR_29ab DR_29Ts DR_30d NR_30O NR_31abS NR_30G

28ab/C30ab 25norC30ab/30ab 29ab/30ab 29Ts/30ab C30d/30ab C30O/30ab 31abS/30ab C30G/30ab

Examples of ion fragmentograms displaying the CEN method’s targeted hopanes and other tri- and pentacyclic triterpanes, rearranged (diasteranes) and regular steranes, and triaromatic steroids are presented in Figure 14.5A–E. Accordingly, definitions of the recommended ratios within each group of biomarkers are listed in ­Tables 14.3–14.5. Tri-aromatic steroid hydrocarbons (derived from steranes) may also be used as diagnostic compounds suitable for oil spill identification as has been described by (Barakat et al. 2002) with respect to terrestrial spilled oil. Aromatic steranes are robust

CEN methodology for oil spill identification

703

Table 14.4  Normative and recommended diagnostic ratios (DR) from rearranged (diasteranes) and regular 14a(H)- and 14b(H)steranes Abbreviation

Compound name

m/z value

27dbS 27dbR 28aaR 29aaS 29bbR 29bbS 29aaR 27bbR 27bbS 28bbR 28bbS 29bbR 29bbS

C27 13b(H),17a(H),20S – Diacholestane (diasterane) C27 13b(H),17a(H),20R – Diacholestane (diasterane) C28 24-methyl-5a(H),14a(H),17a,20R – Cholestane C29 24-ethyl-5a(H),14a(H),17a,20S – Cholestane C29 24-ethyl-5a(H),14b(H),17b(H),20R – Cholestane C29 24-ethyl-5a(H),14b(H),17b(H),20S – Cholestane C29 24-ethyl-5a(H),14a(H),17a(H),20R – Cholestane C27 5a(H),14b(H),17b(H),20R – Cholestane C27 5a(H),14b(H),17b(H),20S – Cholestane C28 24-methyl-5a(H),14b(H),17b(H),20R – Cholestane C28 24-methyl-5a(H),14b(H),17b(H),20S – Cholestane C29 24-ethyl-5a(H),14b(H),17b(H),20R – Cholestane C29 24-ethyl-5a(H),14b(H),17b(H),20S – Cholestane

217 217 217 217 217 217 217 218 218 218 218 218 218

Ratio name

Definition

NR_27dbR/27dbS DR_29aaS DR_29bb NR_27bb DR_C28bb

27dbR/27dbS 29aaS/29aaR 29bb(R+S)/29aa(S+R)* 27bb(R+S)/29bb(R+S)* 28bb(R+S)/(29bb(R+S))*

* As the sterane bbR- and bbS- isomers are often not well-resolved, it is recommended to measure the 27bb(R + S), 28bb(R + S), and 29bb(R + S) double peaks by integrating the whole area and hence, use the obtained areas or the peak heights of the highest peak for generating the relevant diagnostic ratios. For the DR_29aaS ratio, either peak height can be used.

against all kinds of weathering, except photo-oxidation (Bonn-OSINet RR2011, Radovic´ et al. 2014, see also Chapter 20). Six ratios derived from triaromatic steroids are recommended as diagnostic tools. The monoaromatic steroids, typically recorded by m/z 253, are not recommended here as they may coalesce with higher paraffins (above C19) also recorded at this m/z value on GC/MS instruments with low mass resolution (e.g., quadrupoles). Sesquiterpanes may be included as an optional group of biomarkers (Figure 14.5F). Sesquiterpanes include a group of bicyclic (C14–C16 polymethyl-substituted decalins) biomarkers that comprise one of the largest of the terpenoid classes. Thus, sesquiterpanes including drimane and eudesmane are common components of crude oils and ancient sediments. In lighter to middle petroleum products like jet fuel and diesel, where refining processes have removed most of the higher-molecular-weight tetracyclic steranes and pentacyclic triterpanes, the lower-molecular-weight bicyclic sesquiterpanes are generally concentrated and useful. In GC/MS chromatograms, these compounds may be examined by their characteristic fragment ions (m/z 123, 179, 193, and 207) from which highly diagnostic ratios for correlation, differentiation, and

704

Standard Handbook Oil Spill Environmental Forensics

Table 14.5  Normative and recommended diagnostic ratios (DR) derived from triaromatic steroids Abbreviation

Compound name

m/z value

C20TA C21TA SC26TA RC26TA+SC27TA SC28TA

C20-Triaromatic steroid (pregnane derivative) C21-Triaromatic steroid (homopregnane derivative) C26, 20S-Triaromatic steroid (cholestane derivative) C26, 20R- +C27, 20S-Triaromatic steroids C28, 20S-Triaromatic steroid (ethylcholestane derivative) C27, 20R-Triaromatic steroid (methylcholestane derivative) C28, 20R-Triaromatic steroid (ethylcholestane derivative)

231 231 231 231 231

RC27TA RC28TA

Ratio name

Definition

DR_C20TA/C21TA DR_C21TA NR_SC26TA NR_SC28TA NR_RC27TA NR_RC28TA

C20TA/C21TA C21TA/RC26TA+SC27TA SC26TA/RC26TA+SC27TA SC28TA/RC26TA+SC27TA RC27TA/RC26TA+SC27TA RC28TA/RC26TA+SC27TA

231 231

source ­identification of lighter to middle range petroleum products may be acquired (Wang et al., 2005 and Stout and Wang 2008; see Chapter 11). As any other low-boiling compounds, however, the sesquiterpanes are subject to evaporative weathering, a fact that is very valuable for the MS–PW plot, but has to be taken into account before using them for ratio comparison (Table 14.6).

14.8.2.3  Normative ratios (NR) In order to prove a positive match, all chromatograms unaffected by weathering or mixing should be identical and PW plots and ratios should match. To be sure that the ratios used are sufficient to represent the samples involved, a minimum set of normative ratios (indicated by NR) has been defined in the CEN 2012 guidance; these NRs are selected from the ratios proposed in CEN 2012 Annex D and E, which are specified for different, common oil products (Table 14.7). Whereas most of the NRs may be used when crude oils, bunker oils, sludge, and bilge samples are involved, only a limited number may be appropriate for lighter fuel oils (e.g., kerosene, paraffin, diesel, and gas oil) because some of the higher-boiling compounds may not be present in such light-end refined products (see further Chapter 11). Laboratories working according CEN 2012 have to use these ratios as a minimum set for comparison. When ratios for a specific product type cannot be calculated for specific reasons (low S/N, influence by coeluting compounds, etc.), this should be mentioned in the (internal) report.

CEN methodology for oil spill identification

705

Table 14.6  Optional diagnostic ratios (DR) derived from sesquiterpanes Abbreviation

Compound name

m/z value

SES0 SES1 SES2 SES3 SES4 SES5 SES6 SES7 SES8

C14H26-Sesquiterpanes C15H28-Sesquiterpane C15H28-Sesquiterpane C15H28-8b(H)-Drimane C15H28-Sesquiterpane C15H28-Sesquiterpane C16H30-Sesquiterpane C16H30-Sesquiterpane C16H30-8b(H)-Homodrimane

123 123 123 123 123 123 123 123 123

Ratio name

Definition (incl. peak no.)

DR_ SES1/SES3 DR_ SES2/SES3 DR_ SES4/SES3 DR_ SES8/SES3

SES1/SES3 SES2/SES3 SES4/SES3 SES8/SES3

14.8.3  GC/MS – level 2.2: treatment of results 14.8.3.1  Comparison of oil samples using MS–PW plots In the CEN 2012 guideline, the so-called percentage-weathering plot (PW plot) has been introduced as an additional, powerful tool for result evaluation. The CEN 2006 guideline had advised to use only the nonweathered compounds in ratio comparison. But upon using PW plots, more specific differences caused by weathering may be recognized and further contribute to the proof that the weathered sample was identical to the proposed source at the time of the spill. The MS–PW plots can graphically and quickly reveal the effect(s) of weathering on a spill sample relative to its source. To compensate for concentration differences, the spill sample is normalized to the source sample by a stable compound; these include hopane, tetramethyl-phenantrene (T-M-Phe) or phytane, depending on the type of oil (product) involved. This can be achieved in a spreadsheet file in which each spill sample’s targeted analytes’ integrated values (or concentrations) are normalized via Equation (14.1): CN spill /CHopane spill %CN spill = × 100% CN source /CHopane source

(14.1)

In RR2009 the variance of the MS–PW plot was tested by sending blind duplicates of the same solution as samples 1 and 2 to all the participants. Since samples 1 and 2 are definitely the same, the variation of the data-points on the PW plots of sample

706

Standard Handbook Oil Spill Environmental Forensics

Table 14.7  Selected normative diagnostic ratios for different oil product types Light fuel

Ratio

Lube oil

Bilge

HFO

Sludge

Crude oil

V V V

V V V

GC NR_GC-C17/Pr NR_GC-C18/Ph NR_GC-Pr/Ph

V V V

V V V

NR_C17/Pr NR_C18/Ph NR_ Pr/Ph NR_2-MPhe/1MPhe NR_4-MDBT/1MDBT NR_2-MFL/4-MPy NR_B(a)F/4-MPy NR_B(b+c)F/4MPy NR_2-MPy/4-MPy NR_1-MPy/4-MPy NR_Retene/T-Mphe NR_BNT/T-M-phe NR_27Ts/30ab NR_27Tm/30ab NR_28ab/30ab NR_29ab/30ab NR_30O/30ab NR_31abS/30ab NR_30G/30ab NR_27dbR/27dbS NR_27bb/29bb NR_SC26/ RC26+SC27 NR_SC28/ RC26+SC27 NR_RC27/ RC26+SC27 NR_RC28/ RC26+SC27

V V V V

V V V V

V V V V

V V V V

V V V V

V

V

V

V

V

V V V

V V V

V V V

V V V

V V V

V V V

V V V

V V

V V V

V V V

V V V V V V V V V V

V V V V V V V V V V V

V V V V V V V V V V V

V V V V V V V V V V V

V

V

V

V

V

V

V

V

V

MS m/z 85

m/z 192 m/z 198 m/z 216

m/z 234

m/z 191

m/z 217 m/z 218 m/z 231

V V V V V V V V V V

CEN methodology for oil spill identification

707

Figure 14.6  Examples from RR2009 of MS–PW plots for the comparison of samples 1–2, which are from the same injection solution.

1 versus sample 2 merely shows the influence of the analytical method, that is, the analytical variation. Figure 14.6 shows the varying results obtained by four RR2009 participants. Theoretically when two samples are exactly the same all data points should be at 100%, but in practice some variance due to the analysis and the integration will always be visible. MS–PW plot A shows a very low variation around 100% indicating that a minimal analytical variation was achieved at this laboratory. PW plot B shows more earlier eluting compounds (PACs) and also very low deviations from a straight line, but this line is only at 85%. Only the compound that was used for normalization (here T-M-Phe) is at 100%. Of course, in such a situation the first idea must be to check the outlier. It can definitely not be expected that there is a true difference in the two samples in T-M-Phe, if – as in this case – everything else matches (including the results from the three previous analytical steps). Checking outliers is also necessary in PW plot C. This plot shows some more variation of the data points, while most of the data points are in the range of 100–115%. Only a few points are out of this range. PW plot D is definitely not acceptable and the high variation of the data points indicates that the system is not in control.

708

Standard Handbook Oil Spill Environmental Forensics

Generally, the evaluation of the MS–PW plot is the same as the evaluation of the compound ratios: deviations from the 100%-line must be interpreted in the same way as the deviations of the compound ratios (analytical errors or true differences?). For an assessment of the error range of the MS–PW plot, the cumulative standard deviation of the participants of RR2009 has been calculated from the combined data points. To be able to eliminate outliers, the data points were first sorted by participants, starting with the participant exhibiting the lowest variance. The standard deviation (st. dev.) shown for participant 1 in Figure 14.7A is the st. dev. of the data points of participant 1.The st. dev. shown for participant 2 is calculated from all data points of participant 1 and participant 2. The st. dev. shown for

Figure 14.7  (A) Cumulative standard deviation of all data points normalized to T–M–Phe and hopane; (B) cumulative standard deviation of the PW plot for the subselections: normative PACs, normative biomarkers, informative biomarkers, small peaks integrated on height, high peaks integrated on height, and compound groups integrated on area.

CEN methodology for oil spill identification

709

participant 3 is calculated from the results of participant 1, 2, and 3, and so on. The st. dev. shown for participant 20 is calculated from the results of all participants. The results of subselections of the data points where T-M-Phe was used as a normalizing compound were similarly evaluated and are shown in Figure 14.7B. Figure 14.7A and B together show that the cumulative st. dev of the PW plot is almost independent of the type of subselection. The integration on peak height of high peaks (Hhigh) and low peaks (Hsmall) results in the same cumulative st. dev profile. Also, area integrations of PAC alkyl groups result in about the same cumulative st. dev. profile. The flat part of the graphs was considered as an acceptable st. dev. for the MS–PW plot (sPW), that is, values in the range of 7–8. The mean of the combined results is close to the theoretical value of 100%, so the 95% error range will be 100 ± 2 × 7.5, or 100 ± 15%. From a mathematical point of view however, the error range should be 85–118%. Choosing one of the samples as 100% and calculating the results of the other sample relative to it has an effect on the difference to the mean of 100%. The mathematical effect is shown in Table 14.8. It must be pointed out here that the MS–PW plot was a new tool for most participants of RR2009. Some participants did not review or discuss their plots upon reporting, even though these plots were automatically produced upon entering their results into the “automated” Excel spreadsheet provided to each. However, as more participants gained experience in the value of the PW plot, the RR results have improved annually. For RR 2014, only three participants had an sPW of more than 7.5% for specific reasons, while 21 of the 31 participants had an sPW in the range of 2–5%. This serves to demonstrate the improvements in performance that have been achieved through the cooperation and information exchanges among participating laboratories. In addition to the experience gained among participants the use of MS–PW plots has been further improved. Figure 14.8 shows the MS–PW plot of spill 1 versus source 1 from RR2014. The samples have been analyzed in duplicate and after a performance check combined for sample comparison. The data points of Figure 14.6 are sorted by compound group (PACs and biomarkers; normative and informative), while the data points of Figure 14.8 are related to the presumed weathering behavior of these compounds. The “open” marker of the “stable” group indicates compounds firstly reduced in extreme weathering situations except evaporation. The open markers belonging to the weathering processes indicate compounds most sensitive to each process (see Dahlmann and Kienhuis, 2015 and Table 14.8  MS–PW plot data point calculation. Influence on the error range by choosing one of the samples as reference to the other with the applied values in bold Sample 1 peak height Sample 2 peak height

100 100

100 95

100 90

100 85

100 80

100 50

Sample 2/Sample 1 (%) Sample 1/Sample 2 (%)

100 100

95 105

90 111

85 118

80 125

50 200

710 Standard Handbook Oil Spill Environmental Forensics

Figure 14.8  RR2014: MS–PW plot spill 1 versus source 1 normalized to hopane 30ab. Used markers: stable, h or j; biodegradation, ∆; photo-oxidation, o or ; solubility,  or ; others, x and property (sulfur + containing). The “open” marker () for stable compounds indicates compounds firstly reduced by very severe weathering situations except evaporation, whereas the open markers for photo-oxidation and solubility indicate compounds most vulnerable to these forms of weathering to various forms of weathering. (A) Evaporation line. (B) Dissolution line.

CEN methodology for oil spill identification

711

Bonn-OSINet RR 2013 for specific compound information). Upon using these different symbols, the effects of different weathering processes can be directly seen in the MS–PW plots. It is also possible to make use of a “sinuous curved” line to indicate evaporation. Spill 1 of Figure 14.8 is the result of a dissolution experiment performed by Cedre (France). A small layer of crude oil (source 1) was cast on top of a large bottle of seawater and stored in the dark at room temperature for 1 month. During this time, the water was stirred slowly and continually refreshed by a constant but low flow of fresh seawater, in order to extract the more soluble compounds of the crude oil. The left plot in Figure 14.8 shows a “sinuous curved line” that has been adjusted to follow the stable compounds (Me-decalins, sesquiterpanes, pristane and phytane, triterpanes, hopanes, and steranes). Because these compounds are considered very stable, only evaporation can explain their decline at lower retention times. The right plot shows a line that has been adjusted to follow the compounds susceptible to dissolution. As a side effect of the experiment, a reduction of n-C17 (at 86%) and n-C18 (at 87%) was observed, indicating that some biodegradation of these biodegradationsusceptible n-alkanes had taken place, something to be expected after a month at room temperature.

14.8.3.2  Comparison of oil samples using diagnostic ratios If two oil samples are identical, their chemical composition is by definition the same, apart from those changes introduced after the spill as the result of weathering, contamination, mixing, and degradation (Section 14.4). Accordingly, measured ratios between any pair of compounds should also match in identical samples – up to a certain analytically related statistical confidence level. Various ratios between individual compounds or a group of compounds have been described extensively in the geochemical and oil spill literature and these are used to compare oil samples within the CEN guideline. In many cases, such ratios may be referred to as “diagnostic” if they possess the quality to genuinely discriminate different oils. This is often the case with ratios derived from biomarkers, which offer high specificity among different oils (Chapter 4). Other ratios like some based on PACs may or may not possess similar diagnostic significance; especially in refined products, as refinery processes (e.g., thermodynamic equilibria, cracking, and reforming) may have altered any genuine petrogenic PAC distribution (Peters et al., 1992). However, genuinely diagnostic or not, any ratio not influenced by weathering or degradation should still display the same value for identical samples and thus have some “diagnostic” value in a specific spill case and contribute in discriminating and ruling out “nonsource” oils. It is important to realize, however, that the suite of diagnostic PAC and biomarker ratios are neither all-inclusive nor appropriate for all oil spill identification cases. In some instances, it may be prudent to include a certain characteristic feature of the spilled oil that is recognized as particularly diagnostic. In other situations, the abundance of some compounds necessary for determining the recommended diagnostic ratios are below the recommended S/N ratio. Thus, maintaining flexibility in the selection of diagnostic ratios to be used in a specific spill case is very important. However,

712

Standard Handbook Oil Spill Environmental Forensics

reports should provide explanations in the event that certain normative ratios were excluded.

14.8.3.3  Criteria for selecting, eliminating, and evaluating diagnostic ratios In this CEN methodology, the criteria for selecting diagnostic ratios have generally been based on: (1) specificity and diversity, (2) resistance to weathering, and (3) analytical precision/complexity (e.g., ratios calculated using different m/z values are generally not recommended). Basically, only those ratios that can be measured with low variation should be evaluated for comparing candidate sources to spilled oil (Stout et al., 2000, 2001). Peaks with low S/N ratios have an increased variance and should only be used for a visual comparison and not for the comparison of diagnostic ratios. To accommodate both for the limitation in analytical precision and impact of sample heterogeneity, Stout et al. (2001) suggested a protocol by which candidate diagnostic ratios are evaluated in order to identify those that are most useful for further correlation analysis. The evaluation of the diagnostic ratios was conducted by a simple statistical test (RSD) of triplicate analyses of multiple oils in order to identify those ratios that were unaffected by sample heterogeneity or low analytical precision. However, the use of triplicate analyses of one or even multiple samples to determine whether a ratio should be used or not may not be particularly robust and may result in highly varying RSDs among the diagnostic ratios. Another approach was suggested by Faksness et al. (2002a) who applied a student’s t-test on the triplicate analyses of one sample and the 95 and 98% confidence levels as critical limits to determine whether the two ratios matched or not. Again this approach is not very robust, as the triplicate analyses sometimes may result in very low variances for some ratios, and in the end this may result in very small critical differences and hence nonmatching ratios (i.e., false negative or the type I error). For the CEN 2006 and 2012 methodology, a comparable but yet different approach has been applied. It recommends a fixed RSD of 5% for all diagnostic ratios to overcome the variation in critical differences between different ratios and between oil cases. Before applying this approach, however, the laboratory should validate its analytical method by analyzing a crude oil sample at least seven times in order to test that the results comply with the 5% RSD limit. This limit serves as a quality criterion, because methods producing higher RSD values should not be used to analyze and compare oil samples in a forensic context. A weakness with this approach, however, could be that small peaks may have relatively high RSDs. Therefore, in practice, it is generally recommended to analyze some of the samples from a spill case in duplicate and to compare the ratios of the duplicates. When small peaks result in differences larger than the critical difference, these peaks should only be used for a visual comparison. To decide whether or not a particular ratio is sufficiently precise and robust to be used for comparison, the CEN methodology describes two consecutive tests to evaluate diagnostic ratios, as described by the protocol/decision chart shown in (Figure 14.9). The two consecutive tests used for selecting or eliminating diagnostic ratios are: • •

Elimination by means of a signal-to-noise (S/N) tests (Section 14.8.3.5) Elimination by means of the comparison of the duplicate analyses (Section 14.8.3.6)

CEN methodology for oil spill identification

Figure 14.9  Protocol/decision chart for diagnostic ratio evaluation.

713

714

Standard Handbook Oil Spill Environmental Forensics

14.8.3.4  Repeatability limit and critical difference To estimate an acceptable difference between two analytical results, the standard deviation of the analysis method is used (International Organization for Standardization (ISO), 1994a,b). Repeatability (r) is applied as the test method in order to compare individual ratios, assuming that the two samples to be compared are originating from the same source – in this case, a spill sample and a prospective source sample. Repeatability conditions are met when the spill and source samples are analyzed in one series. If the repeatability limit is exceeded, it is beyond reasonable doubt that this assumption is not valid and that the samples are originating from different sources. The repeatability limit of validation is based on standard normal distribution. Determining the repeatability standard deviation sr of an analytical method depends on the quality assurance (QA) system of the laboratory. In general, sr is calculated by analyzing samples relevant for a method at least seven times in one series when a new or revised method is being implemented, and if the method has to be validated. Calculation of the standard deviation or RSD generates r. The repeatability limit r95% (i.e., 95% confidence level) is calculated by multiplying the fixed RSD (sr) with a factor of 2.8 by using equation (14.2): r95% = 2.8 × 5% = 14% (14.2) This implies that when samples are analyzed under repeatability conditions, any ratio with a sr of 5% to be used for the evaluation must not differ more than 14% relatively as critical difference (CD).4 When the absolute difference/mean between a pair of corresponding ratios of two samples to be compared is lower than the critical difference of 14%, then the comparison gives a positive match; if it is higher, the ratios do not match. This test has to be performed for every diagnostic ratio applied in the evaluation. The repeatability limit r95% is a test at the 95% confidence level which implies that 5% or one out of 20 of the pair of ratios slightly above the CD is acceptable without jeopardizing the conclusion. Table 14.9 shows an example of a ratio comparison.

14.8.3.5  Elimination of diagnostic ratios using signal-to-noise (S/N) test Only peaks with S/N > 3–5 should be used for comparing diagnostic ratios. The S/N criterion is based on the method recommended by IUPAC (Ettre, 1993). N is the peak-to-peak value of a part of the noise around a peak and S the peak height. A decision range of 3–5 is given because it may often be difficult to estimate the noise 4

 he repeatability limit (r) is the critical difference between two test results; the associated standard deviaT tion is σ√2. In normal statistical practice, for examining the differences between these two values, the critical difference (CD) used is f times this standard deviation, that is, f × σ√2. For a normal distribution at 95% probability level, f is 1.96 and f × √2 then is 2.77. As the purpose of this guideline is to give some simple “rules of thumb” to be applied by nonstatisticians when examining the test results, a “rounded” value of 2.8 has been suggested instead of f × √2 (International Organization for Standardization, 1994b). Therefore, the repeatability limit r95% is calculated by multiplying sr by 2.8.

CEN methodology for oil spill identification

715

Table 14.9  Example of comparison of diagnostic ratios based on a repeatability limit of 14% Diagnostic ratios (%) Sample A Sample B

Mean (%)

Absolute difference (%)

Relative difference (%)*

Conclusions

48 45

50 50

52 − 48 = 4 55 − 45 = 10

4/50 × 100% = 8 10/50 × 100% = 20

Match Nonmatch

52 55

* The critical difference of 14% is based on the repeatability limit (r95% at 5% RSD), Equation (15.2).

precisely due to the many small peaks present in the chromatographic fine structure of oil samples. As an example, a partial m/z 191 ion chromatogram (C28 [22R] to 27 Tm) of the SINTEF oil mixture is shown in Figure 14.10. The insert within the figure is a part of the ion chromatogram of an injection of clean DCM (Blank injection) over the same retention time section, demonstrating the noise calculation. Normally a part around the target peak should have been taken to calculate the noise. Comparing, however, the mainly electronic noise from the blank injection with the “chemical noise” of the oil sample, shows that the baseline in the oil sample is an ion signal of compounds. In this example, the noise of the blank Nbl = 112. The “compound noise” in the oil sample is Nsample = 4975 and the signal of C28 (22S) Ssample = 20,319 resulting in S/Nsample = 4.1. Compared to the decision range of 3–5, both decisions are valid. Smaller peaks should not be integrated for use in diagnostic ratios, but can still be useful for qualitative visual inspection. Larger peaks should be included in the DR comparison test.

Figure 14.10  Signal to noise calculation. The distance between the highest and lowest peak of a part of the noise is used as N and the peak height as S. S/N is 4.1 for this example.

716

Standard Handbook Oil Spill Environmental Forensics

Even more difficult, however, is the integration of groups of isomer compounds. For example, in de-aromatized (e.g., propeller shaft) lubricating oil, many of the methylated PACs are present at a low concentration showing only the highest peaks of an isomer group. In such cases, the isomer pattern should only be used for visual comparison, and in general, the S/N  > 3–5 criterion should be use as a “rule-ofthumb” to facilitate the elimination of small peaks and groups of peaks from the ratio comparison.

14.8.3.6  Elimination of diagnostic ratios using duplicate analyses Duplicate analyses of two aliquots taken from a homogenous sample or the duplicate injections of the same extract will provide information about the actual performance of the analytical system for the samples involved. To improve the value of this test, it is strongly advised to analyze all samples in duplicate in cases with two or three samples. In cases with many samples, at least 10% of the samples should be analyzed in duplicate. Ideally, duplicate analyses of an extract should result in identical diagnostic ratios, and observed differences can only be caused by analytical variation. As a result of this test, several options are possible according to the repeatability test (Section 14.8.3.4): 1. Ratios of large peaks (i.e., with S/N >> 5) differ more than 14% (Section 14.8.3.4, Equation 14.2), in which case the integration of the peaks involved (retention time and baseline drawing), instrumental conditions and/or injection concentration should be rechecked and the sample (e.g., after cleanup) preferably reanalyzed. 2. Ratios of small peaks (i.e., S/N ≈ 5) differ more than 14%, again the integration of the peaks involved should be rechecked (retention time and baseline drawing); if one or two ratios are still above the critical difference), the peaks involved should only be used for a visual comparison, while if several ratios are still above the critical difference, proceed as with large peaks. 3. All ratios differ less than 14%, then proceed to the next step of the decision chart (Figure 14.9) and further compare the samples involved. Use the mean value of the diagnostic ratios of the samples analyzed in duplicate; the mean value has a lower variance and comes closer to the actual real ratio of the sample.

Generally, duplicate or triplicate analyses make the comparison of diagnostic ratios statistically more robust and reduce the RSD by √n. Hence, the RSD of the mean value (m) of duplicate analyses becomes RSD/√2. For simplicity, however, for the CEN guideline it has been decided to not reduce the repeatability limit of 14% and to not make a distinction between samples analyzed just once or in duplicate.

14.9  Final evaluation and conclusions The final conclusion reached using the CEN guideline should be based on a total evaluation using all available data. The GC–PW and MS–PW plots and comparison of the diagnostic ratios are important parts of the evaluation of the data, however, not alone conclusive. It is important to visually inspect all the chromatograms and

CEN methodology for oil spill identification

717

identify possible characteristic features, and not only to evaluate the measured ratios, before final conclusions are made. In accordance to the flow chart in Figure 14.1, the identification of the source of an oil spill using the CEN methodology should be concluded using one of the four terms: positive match, probable match, inconclusive, or nonmatch. A positive match states that source and spill samples are identical to a high degree of scientific certainty. A visual inspection of the chromatograms (GC/FID) and ion fragmentograms (GC/MS) shows only differences with can be explained unequivocally, for example, by weathering/degradation, and all observed differences between nonweathered diagnostic ratios are below the repeatability limit. A nonmatch applies when the differences between chromatograms and diagnostic ratios cannot be explained by weathering/degradation, and when several pairs of ratios are outside the repeatability limit. If only very small differences (close to the repeatability limit) are observed, if contamination is expected, or if just one pair of ratios is clearly outside the repeatability limit, a probable match is the obvious alternative conclusion. If the total amount of oil in a sample is very low, and consequently there is a higher analytical variance among the diagnostic peaks, which might result in differences between diagnostic ratios (based on repeated analyses of the same sample) that are higher than the recommended repeatability limit. This circumstance would eventually lead to the elimination of so many of the diagnostic ratios from further comparisons that it would render the comparison as inconclusive.

14.10  The CEN methodology in practice: A case study 14.10.1  The spill case During an actual spill event in 2013, a waterborne oil slick was discovered in an inland harbor in The Netherlands. According to observations, on the first day the slick was observed it was blown by a northern wind to an edge of the harbor and the next day it was blown to an opposite edge by a southern wind. On the second day the slick was removed from the water with a skimmer resulting in a total amount of about 5000 L. The case happened in winter time with low sun and a daytime temperature of about 5°C. Three samples were collected from the spill and sent to the laboratory the same day. Analysis with GC/FID indicated that the slick was comprised of diesel fuel. To check for the presence of any small amount of lubricating oil the samples were also analyzed with GC/MS SIM. The ion chromatogram of m/z 191 showed no indication of biomarkers that would indicate the presence of lubricating oil. Based on the analytical results, the three spill samples were found to be very similar, indicating that the slick was probably from a single source of diesel fuel. Based on this preliminary information, 6 days later 14 prospective source samples from the diesel fuel tanks of 11 vessels were collected and sent to the analytical laboratory for oil spill identification.

718

Standard Handbook Oil Spill Environmental Forensics

14.10.2  Sample preparation As noted previously, sample preparation should be done in the simplest way to prevent changes in composition. Therefore in this case the thin layer of floating oil on water from two of the spill samples were collected in small vials to be able to dilute the floating oil. This prevents the extracting of the whole samples, which may introduce contamination. Spill samples 1 and 3 contained a thin layer of oil on water, while spill sample 2 contained a layer of about 3 cm of diesel on water. Sample 1 and 2 were collected in a bottle from the water surface, while sample 3 was from a tank after collection of the slick by the skimmer. The water part of spill 3 was colored grey and contained a lot of particles. From spill 1 and 3 about 6 mL of the top layer was collected with a Pasteur pipette in 8 mL vials. After separation of the layers 20 mL of spill 1 was diluted in 20 mL DCM and 20 mL of spill 3 in 4 mL DCM. 1 mL of the spill 3 solution was cleaned on a column made of a Pasteur pipette filled with about 1 gram of Florisil and topped by about 1 g of dry sodium sulfate (Albaigés et al., 2015a). The column was eluted with DCM up to a final volume of 5 mL, resulting in the same injection concentration as the other samples. From spill 2 and all 14 source samples, 20 mL was diluted in 20 mL DCM. In order to further evaluate the results using COSIweb (see Chapter 15) all samples were analyzed with GC/FID and GC/MS within a single analytical sequence, with spill sample 2 and source samples 6 and 13 being analyzed in duplicate. The stability of the sequences was checked by analyzing a standard of n-alkanes before and after the GC/FID samples and a standard of Brent crude oil before and after the GC/MS SIM analysis section of the sequence. The ratios n-C10/n-C20 and n-C40/n-C20 of the first and last standard together with the ratio n-C20/n-C40 between the first and last standard were checked by means of a control chart. In the same way, ratios between three compounds (dimethylnaphthalenes, 3-methylphenanthrene and C30-hopane) were used to check the stability of the GCMS SIM analysis. The results indicated that the sample results from the analytical sequences could be evaluated. After uploading the GC/FID and GC/MS chromatograms to COSIweb the duplicates were checked. Comparison to the COSIweb database showed that the differences between the duplicates were far below the criteria for the PW plots and the diagnostic ratios. The three spill samples were then compared to each other. Spill samples 1 and 3 were more evaporated than spill sample 2, which seems reasonably attributed to the thickness of the oil in the bottles received. Despite different levels of evaporation, a positive match could be concluded between the three spill samples – confirming the three samples were derived from a single source. As such, the results of spill sample 2 (i.e., the thick layer of 3 cm on water) were used to represent the spill during comparison to the prospective source oils using the CEN methodology.

14.10.3  Level 1 – GC/FID screening A visual inspection of the GC/FID chromatograms revealed that all spill samples and four (of 14) source samples contained some biodiesel. See Figure 14.2 for spill 2 and source 2 and Figure 14.11 for spill 2 and sources 1, 3, and 4. Several FAME peaks are visible and dominated by C18:1 eluting just before n-C21 (DeMello et al., 2007). FAMEs can be analysed by m/z 55, 67, 74 and 79, but these are not mentioned in CEN 2012. So the presence of FAMEs was confirmed by GC/MS

CEN methodology for oil spill identification

719

Figure 14.11  GC/FID chromatograms of extracts of spill 2 and sources 1, 3, and 4.

ions m/z 83 (see Figure 14.14) and 234 (see Figure 14.13). Based on these results and combined with the results of COSIweb, it was decided to manually compare the spill samples with the four source samples containing biodiesel for the following reasons: 1. The presence of biodiesel in the source samples of a case, but not in the spill samples would not directly result in the exclusion of source samples with biodiesel, because the weathering of biodiesel in spill cases is not described in CEN 2012. Oppositely, however, when biodiesel is present in all spill samples at the same level, it is very unlikely that the source samples containing no biodiesel could possibly be the source of the spill samples. 2. COSIweb has been developed for crude oil and HFO samples, but can also be used for diesel and bilge samples. All the normative ratios for light fuels (see Table 14.2) are available for comparison. The database shows a high correlation between the spill samples and source sample 2. Several of the other source samples show a reasonable correlation with some minor differences. At present, COSIweb is missing some early eluting compounds like the sesquiterpanes. In order to improve the assessment the samples were integrated manually and the values compared by means of a spreadsheet file.

The GC–PW plots of spill sample 2 compared with the four source samples containing biodiesel are shown in Figure 14.12. The GC–PW plot of spill sample 2 compared with source 2 shows a very high similarity. The “evaporation line” follows the sinuous curved pattern up to n-C14, while the 100% range after n-C14 shows a very low variance around 100%. The other three source samples’ plots show more differences in the alkane patterns and small to larger differences for pristane and phytane. This is confirmed by the isoprenoid ratio calculations. Table 14.10 shows that all differences relative to the mean in % are lower than the repeatability limit of 14% (see Table 14.9), but also that the differences between source sample 2 and spill sample 2 are extremely low. Based on the GC/FID results,

720 Standard Handbook Oil Spill Environmental Forensics

Figure 14.12  GC–PW plots of extracts of spill sample 2 with the four source samples containing biodiesel.

CEN methodology for oil spill identification

721

Table 14.10  Comparison of the ratios of pristane and phytane of spill sample 2 with the source samples 1–4 using the repeatability limit Diagnostic ratio n-C17/Pr n-C18/Ph Pr/Ph

Spill 2

Source 1

Source 2

2.14 2.44 1.44

2.05 2.24 1.38

2.16 2.45 1.45

Source 3 2.02 2.20 1.36

Source 4 1.87 2.26 1.49

Relative difference/mean (%) of spill 2 with the source samples n-C17/Pr n-C18/Ph Pr/Ph

4.3 8.4 4.1

1.1 0.6 0.6

5.4 10.3 5.4

13.1 7.8 3.8

none of the four source samples containing biodiesel could be eliminated as possible sources, and all four samples required further assessment at level 2.

14.10.4  Level 2 – GC/MS fingerprinting 14.10.4.1  Level 2.1 – visual inspection and elimination The evaluation was continued to level 2, GC/MS fingerprinting, according to the flow chart in Figure 14.1. Before integration, the ion-chromatograms were compared visually to check for specific features (see Figure 14.13).

Figure 14.13  Ion chromatograms of Brent crude oil and spill 2 for (A) m/z 234 (C4-phenantrenes), retene and benzonaphthothiophene (BNT), and (B) m/z 216 (C1-fluoranthenes/pyrenes).

722

Standard Handbook Oil Spill Environmental Forensics

As is shown in Figure 14.13, the compounds 2-methylfluoranthene (2-MFL), benzo[a]fluorene (B[a]F), benzo[b + c]fluorene (B[b + c]F), retene and BNT used to calculate numerous normative ratios for light fuel oil (Table 14.2) were not present in the spill samples, or were at a low level (S/N < 3–5). In addition to the remaining normative ratios, other informative compounds and ratios were used in order to improve the number of features available for numerical assessment.

14.10.4.2  Level 2.2 – MS–PW plots and diagnostic ratios Figure 14.14 shows the comparison of spill sample 1 with the combined values of the duplicate analyses of spill sample 2. The ion chromatogram of m/z 83 has been added to show the usefulness of this ion to indicate the presence of FAME’s. The MS–PW plot indicates that spill sample 1 is more evaporated than spill sample 2. For example, the C1-decalins (group integrated by area and eluting at 13 min) in spill sample 1 were reduced to 38% relative to spill sample 2 (Figure 14.14, MS–PW plot). The sesquiterpanes (several data points eluting between 20–24 min) are at the 100% level, indicating they have not been affected by evaporation. In between the decalins and sesquiterpanes, which are very resistant against all types of weathering except evaporation, naphthalene and the C1- naphthalenes were reduced in the spill 1 sample by 40 and 83%, respectively. These 2-ring aromatic compounds dissolve into water more readily relative to the other PAHs, and therefore their reduction could be attributed to dissolution. However in this case, these compounds fall along the evaporation line indicating their reduction is likely caused by evaporation only. The other data points of the MS plot are at the 100% level, indicating a high similarity

Figure 14.14  m/z 83 ion chromatogram of spill 2. MS–PW plot of spill sample 1 relative to spill 2 normalized to phytane. Ratio plot showing the difference/mean (%) for the ratios of spill 1 compared to spill 2.

CEN methodology for oil spill identification

723

Figure 14.15  MS–PW plot normalized to phytane of spill sample 2 relative to the source samples 1–4.

between the two spill samples. This is confirmed by the ratio graph on the right side of Figure 14.14, which shows that all ratio differences are below (mostly far below) the critical value of 14%, further confirming the high similarity of these spill samples. Figure 14.15 shows the MS–PW plots of spill sample 2 compared with the four source samples that also contained biodiesel. The MS–PW plot of spill 2 – source 2 is very similar to the MS–PW plot of the two spill samples (Figure 14.14) indicating a good similarity between both samples. The plots for the other three source oils, however, are markedly different, wherein the data points are scattered. This degree of scatter indicates the source 1, 3, and 4 oils cannot be the source of the spill oil; that is, they appear to be nonmatches. The apparent match of source 2 and apparent nonmatches for the other source oils revealed in the PW-MW plots is confirmed by the ratio graphs shown in Figure 14.16. The graphs of the source sample 1, 3, and 4 show several ratio differences exceed the 14% line, sometimes significantly higher. However for source sample 2, all of the ratio differences are below, mostly far below, the 14% line indicating that no larger ratio differences exist (other than those attributable to the analysis and data evaluation).

14.10.5  The CEN methodology in practice: conclusions In practice, it is typically easier to conclude a nonmatch than a positive match. In this case study, source samples 5–14 could all be quickly excluded by the absence of biodiesel, which was present in all three spill samples. Additionally, COSIweb showed a low correlation existed between the spill samples and source samples 5–14. The manual evaluation of spill 2 compared with source samples 1, 3, and 4 revealed marked

724

because of the thick layer of oil on water for spill 2 and the combination of the duplicate analyses of spill 2.

Standard Handbook Oil Spill Environmental Forensics

Figure 14.16  Ratio graphs between spill sample 2 and the source samples 1–4. Spill 2 is indicated here by “water thick + water thick”

CEN methodology for oil spill identification

725

differences existed in the GC–PW plots, MS–PW plots, and ratio graphs, clearly indicating these three source oils were nonmatches to the spill oils. In order to prove a positive match existed between the spill samples and source 2 with a high degree of scientific certainty, all measures were compared and found to be identical. The GC–PW plots, MW–PW plots, and ratio graphs indicated source 2 was a positive match to the spilled oils. However, before reaching this conclusion, the GC/FID and all GC/MS ion chromatograms were further compared visually and no differences were found that could not be explained unequivocally, for example, by weathering. Based on all of these results, a positive match was concluded between the three spill samples and source sample 2.

14.11 Summary The CEN methodology for oil spill identification presented herein is based on a tiered approach including GC/FID or GC/MS full-scan screening of all spill and prospective source samples available (Level 1.1), testing weathering with GC–PW plots, and selected ratios (Level 1.2). After exclusion of any clearly nonmatching source samples followed by GC/MS fingerprinting (Level 2.1), from which MS–PW plots and diagnostic ratios of selected PAHs and biomarkers are derived, diagnostic ratios are selected and evaluated on the basis of their analytical variability and changes due to weathering, and correlation of spill and candidate oil samples based on those diagnostic ratios that can be precisely measured and are resistant to weathering effects (Level 2.2). By a statistical treatment of the ratios, and an overall assessment of results from all analytical levels, the oil spill identification using this methodology can, with a high degree of scientific certainty, be concluded to be one of four operational and technically defensible terms: Positive match, probable match, inconclusive, or nonmatch. The method uses a suite of quality checks: Sample cleanup to protect the GC column and to facilitate good analytical results The analysis of a reference standard (e.g., a reference crude oil, such as Brent crude) around the sequence, with quality checks to test the overall performance of the instrumentation • Minimum resolution of selected peak pairs • Duplicate analyses of at least two or 10% of the samples of a case to test the integration and the actual variance of the sequence and the samples involved • Variance limitations for the MS–PW plot (st. dev. < 7.5%) and the ratio comparison (RSD < 5%) • •

The CEN methodology has, at present, been implemented by at least 35 forensic oil spill laboratories worldwide and, as such, has been used in connection to many oil spill identification cases in, for example, Denmark (Hansen et al., 2002), The Netherlands’ many inland cases but also, for example, the Tricolor incident in the British Channel (Dahlmann and Kienhuis 2015), the Erika incident in France, and the Prestige incident in Spain and France (Bonn-OSINet RR 2011, Guyomarch, 2002; Albaigés et al, 2015b, Chapter 22), the Deepwater Horizon spill (e.g., Bacosa et al., 2015;

726

Standard Handbook Oil Spill Environmental Forensics

Radovic´ et al., 2014; Kolian et al., 2015), the Hebei Spirit spill (Yim et al., 2011) and in numerous small oil spills. The methodology has been demonstrated to be a strong technically defensible tool capable of differentiating among qualitatively similar oils from a spill and all available candidate source oils.

Acknowledgment Bonn-OSINet works as an open source community in which both governmental and commercial laboratories work together with the restriction being that only laboratories actually working with oil spill identification can participate. Coordination of Bonn-OSInet and the use of COSIweb is facilitated by the Bonn-Agreement and the governments of The Netherlands (Ministry of Infrastructure and the Environment) and Germany (Federal Ministry of Transport and Digital Infrastructure). Participation in the Bonn-OSINet annual Round Robin exercises is at present without a fee and only the traveling costs have to be paid by the participants to join the annual meetings. Many participants have contributed by preparing and sending out the RR test samples and/or hosting the annual meetings. Many participants have also contributed by sharing information and knowledge to improve the method.

References Aeppli, C., Nelson, R.K., Radovic, J.R., Carmichael, C.A., Reddy, C.M., 2014. Recalcitrance and degradation of petroleum biomarkers in Deepwater Horizon oil upon abiotic and biotic natural weathering. Environ. Sci. Technol. 48, 6726–6734. Albaigés J., Kienhuis P.G.M., Dahlmann G., 2015a. Oil spill identification. In: Fingas, M. (Ed.), Handbook of Oil Spill Science and Technology. Wiley Online Library. pp. 165–203, Chapter 6. doi: 10.1002/9781118989982.ch6. Albaigés J., Bernabeu A., Castanedo S., Jiménez N., Morales-Caselles C., Puente A., Viñas L., 2015b. The prestige oil. In: Fingas, M. (Ed.), Handbook of Oil Spill Science and Technology. Wiley Online Library. pp. 515–546, Chapter 22. doi: 10.1002/9781118989982.ch22. Bacosa, H.P., Erdner, D.L., Liu, Z., 2015. Differentiating the roles of photo-oxidation and biodegradation in the weathering of Light Louisiana Sweet crude oil in surface water from the Deepwater Horizon site. Mar. Pollut. Bull. 95 (1), 265–272. Barakat, A.O., Qian, Y., Kim, M., Kennicutt, II, M.C., 2002. Compositional changes of aromatic steroid hydrocarbons in naturally weathered oil residues in the Egyptian Western Dessert. Environ. Forensics 3, 219–225. Bonn-OSINet, 2005. The Bonn agreement oil spill identification network of experts. See http:// www.bonnagreement.org/osinet/oil-spill-identification. (The annual round robin reports can be downloaded here.) CEN/TR 15522-1, 2006a. Oil Spill Identification – Waterborne Petroleum and Petroleum Products – Part 1 Sampling, 23 p. CEN/TR 15522-2, 2006b. Oil Spill Identification – Waterborne Petroleum and Petroleum Products – Part 2 Analytical Methodology and Interpretation of Results, 109 p. (No longer available on-line; superseded by CEN 2012). CEN/TR 15522-2, 2012. Oil Spill Identification – Waterborne Petroleum and Petroleum Products – Part 2 Analytical Methodology and Interpretation of Results, 138 p.

CEN methodology for oil spill identification

727

Christensen, J.H., Hansen, A.B., Tomasi, G., Mortensen, J., Andersen, O., 2004. Integrated methodology for forensic oil spill identification. Environ. Sci. Technol. 38, 2912–2918. Dahlmann, G., 2003. Characteristic Features of Different Oil Types in Oil Spill Identification. Berichte des BSH, Germany. 48 p. Available from: http://www.bsh.de/de/Produkte/ Buecher/Berichte_/Bericht31/Bericht31.pdf. Dahlmann, G., Kienhuis P.G.M., 2015. Oil spill sampling and the Bonn-oil spill identification network: a common method for oil spill identification. In: Carpenter, A. (Ed.), Oil Pollution in the North Sea. The Handbook of Environmental Chemistry. Springer, Berlin Heidelberg, New York. Daling, P.S., Faksness, L.-G., Hansen, A.B., Stout, S., 2002. Improved and standardized methodology for oil spill fingerprinting. Environ. Forensics 3, 263–278. DeMello, J.A., Carmichael, C.A., Peacock, E.E., Nelson, R.K., Samuel Arey, J., Reddy, C.M., 2007. Biodegradation and environmental behaviour of biodiesel mixtures in the sea: an initial study. Mar. Pollut. Bull. 54, 894–904. Ettre, L.S., 1993. Nomenclature for chromatography, IUPAC Recommendations 1993, Section 4. 3. Pure Appl. Chem. 65 (4), 819–872. Faksness, L.G., Weiss, H., Daling, P.S., 2002a. Revision of the Nordtest methodology for oil spill identification – technical report; SINTEF report STF66 A01028. Faksness, L.G., Daling, P.S., Hansen A.B., 2002b. CEN/BT/TF 120 oil spill identification. Summary report: Round Robin Test Series B; SINTEF report STF66 A02038. Faksness, L.-G., Daling, P.S., Hansen, A.B., Kienhuis, P., Duus, R., 2005. New guidelines for oil spill identification of waterborne petroleum and petroleum products. Proceedings of the 28th Arctic and Marine Oilspill Program (AMOP) Technical Seminar, Calgary, Canada, June 7–9, 2005. Environment Canada, Ottawa, vol. 1, pp. 183–202. Fuller, S., Spikmans, V., Vaughan, G., Guo, C., 2013. Effects of weathering on sterol, fatty acid methyl ester (FAME), and hydrocarbon profiles of biodiesel and biodiesel/diesel blends. Environ. Forensics 14 (1), 42–49. Guyomarch, J., 2002. Identification du fuel du prestige – 2002; CEDRE Report No. GC.02-15; 2002 (in French). Hansen, A.B., Avnskjold, J., Rasmussen, C.A., 2002. Application of PAH and Biomarker Diagnostic Ratios in Forensic Oil Spill Identification by the Revised Nordtest Methodology, Oil and Hydrocarbon Spills III: Modelling, Analysis and Control. WIT Press, Southampton, UK, pp. 59–66. International Organization for Standardization (ISO) 5725-2, 1994a. Accuracy (trueness and precision) of measurement methods and results; Part 2 Basic method for the determination of repeatability and reproducibility of a standard measurement method. ISO TC69/SC6. International Organization for Standardization (ISO) 5725-6, 1994b. Accuracy (trueness and precision) of measurement methods and results; Part 6 Use in practice of accuracy values. ISO TC69/SC6. Kolian, S.R., Porter, S.A., Sammarco, P.W., Birkholz, D., Cake, Jr., E.W., Subra, W.A., 2015. Oil in the Gulf of Mexico after the capping of the BP/Deepwater Horizon Mississippi Canyon (MC-252) well. Environ. Sci. Pollut. Res. 22 (16), 12073–12082. Nordtest, 1991. Oil Spill Identification, NT Chem 001, second ed. Nordic Innovation Centre, Espoo, Finland. Peters, K.E., Scheuerman, G.L., Lee, C.Y., Moldowan, J.M., Reynolds, R.N., Pena, M.M., 1992. Effects of refinery processes on biological marker. Energ. Fuel. 6 (5), 560–577. Peters, K.E., Walters, C.C., Moldowan, J.M., 2005. The Biomarker Guide, second ed. Biomarkers and Isotopes in the Petroleum Exploration and Earth History, vol. 2. Cambridge University Press, Cambridge, UK. pp. 475–1155.

728

Standard Handbook Oil Spill Environmental Forensics

Radovic´, J.R., Aeppli, C., Nelson, R.K., Jimenez, N., Reddy, C.M., Bayona, J.M., Albaigés, J., 2014. Assessment of photochemical processes in marine oil spill fingerprinting. Mar. Pollut. Bull. 79 (1–2), 268–277. Shang, D., Kim, M., Haberl, M., 2014. Rapid and sensitive method for the determination of polycyclic aromatic hydrocarbons in soils using pseudo multiple reaction monitoring gas chromatography/tandem mass spectrometry. J. Chromatog. A 1334, 118–125. Stout, S.A., Wang, Z., 2008. Diagnostic compounds for fingerprinting petroleum in the environment. In: R.E., Hester, R.E., Harrison, R.M. (Eds.), Environmental Forensics, Issues in Environmental Science and Technology. The Royal Society of Chemistry, London. Special Publication No. 26. pp. 54–104. Stout, S.A., Naples, W.P., Uhler, A.D., McCarthy, K.J., Roberts, L.G., Uhler, R.M., 2000. Use of ouantitative biomarker analysis in the differentiation and characterization of spilled oil. Paper prepared for the SPE International Conference on Health, Safety, and the Environment, Stavanger, Norway, June, 26–27, 2000. Stout, S.A., Uhler, A.D., McCarthy, K.J., 2001. A strategy and methodology for defensibly correlating spilled oil to source candidates. Environ. Forensics 2, 87–98. Stout, S.A., Uhler, A.D., McCarthy, K.J., Emsbo-Mattingly, S., 2002. Chemical fingerprinting of hydrocarbons. In: Murphy, B., Morrison, R.D. (Eds.), Introduction to Environmental Forensics. Academic Press, San Diego, CA. pp. 137–260. Strøm-Kristiansen, T., Lewis, A., Daling, P.S., Hokstad, J.N., Singsaas, I., 1997. Weathering and dispersion of naphthenic, asphaltenic and waxy crude oils. Proceedings of the 1997 International Oil Spill Conference, Trondheim, Norway, pp. 631–636. Wang, Z., Fingas, M.F., 2003. Development of oil hydrocarbon fingerprinting and identification techniques. Mar. Pollut. Bull. 47 (9-12), 423–452. Wang, Z., Fingas, M., Page, D.S., 1999. Oil spill identification. J. Chromatog. A 843, 369–411. Wang, Z., Yang, C., Fingas, M., Hollebone, B., Peng, X., Hansen, A.B., Christensen, J.H., 2005. Characterization, weathering, and application of sesquiterpanes to source identification of spilled lighter petroleum products. Environ. Sci. Technol. 39, 8700–8707. Wang, Z., Yang, C., Brown, C., Hollebone, B., Landriault, M., 2008. A case study: distinguishing pyrogenic hydrocarbons from petrogenic hydrocarbons. International Oil Spill Conference Proceeding, no. 1, May 2008, pp. 311–320. Weiss, H.M., Wilhelms, A., Mills, N., Scotchmer, J., Hall, P.B., Lind, K., Brekke, T., 2000. The Norwegian Industry Guide to Organic Geochemical Analyses (NIGOGA), fourth ed. Norsk Hydro, Statoil, Geolab Nor, SINTEF Petroleum Research, and the Norwegian Petroleum Directorate. 102 p. Available from: http://www.npd.no/engelsk/nigoga/nigoga.pdf. Yang, C., Wang, Z., Liu, Y., Yang, Z., Li, Y., Shah, K., Zhang, G., Landriault, M., Hollebone, B., Brown, C., Lambert, P., Liu, Z., Tian, S., 2013. Aromatic steroids in crude oils and petroleum products and their applications in forensic oil spill identification. Environ. Forensics 14 (4), 278–293. Yim, U.H., Ha, S.Y., An, J.G., Won, J.H., Han, G.M., Hong, S.H., Kim, M., Jung, J.-H., Shim, W.J., 2011. Fingerprint and weathering characteristics of stranded oils after the Hebei Spirit oil spill. J. Hazard. Mater. 197, 60–69.