Journal of Archaeological Science 145 (2022) 105643 Contents lists available at ScienceDirect Journal of Archaeological Science journal homepage: www.elsevier.com/locate/jas Bayesian mixing models as a tool to explore Bronze Age bitumen trade from Tell Lashkir (Erbil, Iraq) Adrià Breu a, *, Antoni Rosell-Melé b, c, Miquel Molist a, Anna Bach-Gómez a a Grup de Recerca en Arqueologia del Mediterrani i el Proper Orient (GRAMPO), Departament de Prehistòria, Universitat Autònoma de Barcelona, Bellaterra, Catalonia, Spain b Institute of Environmental Science and Technology (ICTA-UAB), Universitat Autònoma de Barcelona, Bellaterra, Catalonia, Spain c Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Catalonia, Spain A R T I C L E I N F O A B S T R A C T Keywords: Bitumen from north-eastern Mesopotamian seeps was exchanged across the Tigris and the Euphrates in antiquity. Bronze age mesopotamia Nonetheless, recent research proposed a decline in its trade during the Bronze Age alongside the boost of central Tell lashkir Mesopotamian (Hit) seeps trade. This paper presents new provenance studies from ancient bitumen recovered at Bitumen trade Tell Lashkir, a 4th-2nd millennium BC settlement between the lower and upper Zab Tigris tributaries. The novel Petroleum biomarkers Bayesian mixing models application of Bayesian mixing models to biomarker and isotopic data has narrowed down the identification of the bitumen putative sources and improved the apportionment of mixtures. This placed Tell Lashkir within the pathway of bitumen active trade routes in the Bronze Age. Thus, the population inhabiting Tell Lashkir participated in exchange networks which reached southern Mesopotamia, and traded with natural resources most probably originating from central and north-eastern Mesopotamia. From a methodological standpoint, the appraisal of the Bayesian mixing models in our study with modern and archaeological reference samples has provided a better understanding of the strengths and limitations of such an approach to apportion bitumen sources from several Bronze age territories. 1. Introduction regions involved in the long distance trade of bitumen were south- western Iran and central and northern Mesopotamia (Connan and Van In ancient times, bitumen was a valuable commodity used for its de Velde, 2010). Nevertheless, during the Uruk period, bitumen from Hit stickiness, ignition, and waterproofing properties. A testimony of its (central Mesopotamia) also became prevalent across the Euphrates, thus importance for the first states in Mesopotamia was its use in monu­ possibly replacing previous trade networks at a time when the con­ mental architecture, shipbuilding, water management, art pieces, magic struction of monumental architecture would have demanded increasing and medicine (Forbes, 1964). The term for bitumen and its extensive quantities of this product (Connan and Van de Velde, 2010; Schwartz derived vocabulary appear in cuneiform tablets attesting both com­ and Hollander, 2016). Texts from the Ur III period and molecular ana­ mercial exchanges and political fights for its control (Stol, 2012). This lyses suggest that, through the third millennium BC., Madga bitumen natural resource played a significant role in the economic and political from north-eastern Mesopotamia still reached southern cities (Connan developments of the Bronze Age as it was not only exchanged in either and Van de Velde, 2010). Nevertheless, the scarcity of molecular data raw or processed form, but its use in boat construction facilitated trade across the Tigris in the Bronze Age (Connan et al., 2013, 2018) prevents itself. The molecular and isotopic analyses of ancient bitumen samples obtaining a clear picture of the extension of the bitumen trade networks have revealed that this product circulated across south-west Asia from at the moment of the formation of the first empires in history. Did set­ early prehistory to medieval times (Connan, 1999; Connan and Van de tlements near the Tigris bank or across its tributaries use bitumen pre­ Velde, 2010). This is thanks to the chemical characterization of putative senting the same geochemical fingerprints as those from the southern sources (i.e., oil seeps) and the identification and comparison of their cities which they allegedly served? Were bitumen types more frequent in geochemical fingerprints. settlements across the Euphrates or the Persian Gulf present in the As suggested by recent studies in the Ubaid period, the three main northern Tigris? The range of uses bitumen had in north-eastern * Corresponding author. E-mail address: adria.breu@uab.cat (A. Breu). https://doi.org/10.1016/j.jas.2022.105643 Received 9 October 2021; Received in revised form 15 June 2022; Accepted 10 July 2022 Available online 20 July 2022 0305-4403/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). A. Breu et al. Journal of Archaeological Science 145 (2022) 105643 Mesopotamia, where the Neo-Assyrian empire would develop during the possible existence of mixtures from different areas; its ability to suggest Iron Age, might have been different, as building materials such as stone the possible broad location of unknown sources, and, to treat both were more readily available than in the south (Forbes, 1964; Moorey, reference data and results as probability distributions carrying a degree 1994). This, in turn, could have affected the preferred types of bitumen of uncertainty (Fernandes et al., 2018). To strengthen these claims, the throughout the region. validity of the model deployed in this study was tested using a signifi­ To contribute further to the current understanding of Bronze Age cant corpus of published modern references from a single point of origin bitumen trade across the Tigris this paper reports the study of the and previously identified archaeological samples. bitumen remains recovered at Tell Lashkir (hereafter TL). This is an In consequence, the four objectives of this study are: archaeological site located roughly 10 km east from ancient Arbela and 100 km north from modern Kirkuk. Our study is the first that tackles the a) To test the potential of Bayesian models to better distinguish and origins of ancient bitumen from the land around the lower and the upper apportion bitumen remains with different geographic origins. Zab rivers, and offers the opportunity to understand better the structure b) To assess the variability of potential bitumen fingerprint types within of the exchange networks which supported the demand of this natural TL. resource by the bustling Sumerian and Akkadian cities and states. c) To study the similarities between TL bitumen and archaeological To identify the most probable sources of bitumen, we undertook bitumen from other contemporary sites. qualitative and semi-quantitative molecular and isotopic analyses of d) To estimate the possible original source of TL’s archaeological black amorphous visible residues inside pottery vessels using gas chro­ bitumen. matography coupled to Mass Spectrometry (GC-MS), and Elemental Analysis coupled to Isotope Ratio Mass Spectrometry (EA-IRMS). For the 2. Materials and methods first time in the study of ancient bitumen, the identification of putative sources and its comparison with other published results was performed The site of Tell Lashkir (Fig. 1), is a chronologically long lived set­ through Bayesian mixing models. To date, these have been successfully tlement with indices of Neolithic presence and clear Uruk, Ninevite 5 applied to quantify the proportions of different foodstuffs in archaeo­ (sectors 3 and 6), Middle Bronze Age and Iron Age phases (Gómez-Bach logical pottery residues (Lucquin et al., 2018; Courel et al., 2020) and to et al., 2018, 2019; Molist et al., 2019a; 2019b). Excavations at TL started study the composition of past human diets (Fernandes et al., 2015). This in 2015 thanks to a collaboration between the Autonomous University of statistical tool was selected because of its capability to evaluate the Barcelona, The Directorate of Antiquities of the Kurdistan regional Fig. 1. Location of Tell Lashkir (star), other archaeological sites mentioned in the text (rhomb) and the approximate regions of origin of the modern bitumen references used in the Bayesian models. Source of the data: (Connan and Nishiaki, 2003; Connan et al., 2005, 2013, 2018). 2 A. Breu et al. Journal of Archaeological Science 145 (2022) 105643 government and the department of archaeology in the Salahaddin Uni­ Conflo III interface (EA-IRMS). The δ13C measurements are reported versity (Molist, 2015). Since the start of the excavation, significant relative to the V-PDB standard. quantities of bitumen-bearing potsherds were detected in the site (Breu et al., 2019). Hitherto, four excavation campaigns have certified the 2.2. Statistical analyses presence of bitumen in almost all sectors, especially at the Ninevite 5 period, at the end of the Bronze Age and in the Iron Age (Molist et al., A series of non-weighted concentration-independent Bayesian mix­ 2019b). ing models were built using the FRUITS v3.1beta software (Fernandes This paper presents the results of an initial study which included a et al., 2018) to quantify the relative amounts of different bitumen types total of 12 samples. Five samples were recovered from Sector 1; two to a hypothetic mixture defined by specific geochemical ratios. An un­ samples were from sector 3; four samples from sector 6; and one sample derlying assumption of this approach is that the source of any com­ from sector 11. They were all amorphous black visible residues attached pounds extracted from the visible amorphous residues could only be the to the interior of pottery sherds (see images of the samples in supple­ same components in the bitumen references. In all models, a Markov mentary materials 1). Regarding their context of discovery, Sectors 1, 3 Chain Monte Carlo (MCMC) method using Gibbs sampling and the and 11 presented walls and architectural features associated with do­ Metropolis-Hastings algorithm was run using the BUGS software. It mestic areas. In Sector 1, placed at the start of the Middle Bronze Age performed 10000 burn-in steps and, afterwards, ran for 40000 itera­ (MBA), four samples were taken from a single large building presenting tions. Following Fernandes et al., (2018) the general minimum uncer­ several storage areas constructed after 2059–1900 2σ cal BC and one tainty of the model was set at 0.001 to account for possible unknown sample originated from layers dated after this date. Sector 3 (Ninevite 5, sources of uncertainty. 2887-2636 2σ cal BC, Molist et al., 2019a), containing up to six archi­ Several Bayesian mixing models were deployed to assess data from tectural phases, presented vestiges of earthen architecture related to this study. The first model detected variability within bitumen samples rectangular buildings. In this sector, samples were collected from the from our site; the second evaluated the similarity of the geochemical most recent occupation phase. Finally, samples from sector 6 were ob­ fingerprints from TL to other contemporary archaeological sites, and, tained from layers containing exterior areas related to a nearby human the third explored the possible geographic location of TL’s bitumen installation placed in the first half of the third millennium cal. BC (Early, sources. In all cases, the model outputs were interpreted as the per­ Middle Ninevite 5, Early Jezirah I-II, Amuq H and G (Molist et al., centage of the total bitumen by weight and no informative priors were 2019b). used. In the first model, the 8 measured geochemical ratios (Ts/Tm, Ts/ 2.1. Extraction and analysis of archaeological bitumen 30αβ, Tm/30αβ, 29αβ/30αβ, 30βα/30αβ, 31αβS/30αβ, the 30G/30αβ and 29ββ/29αα) were used as proxies and were described by their me­ Roughly 0.5 g of the black amorphous crusts adhered to internal dian and their standard deviation. Reference values for different vessel walls were removed and ground with a pestle and mortar. The bitumen types were selected from representative samples in TL after resulting samples were extracted with ca. 5 ml of dichloromethane. After statistical evaluation. This included a Principal Component Analysis 15 min of sonication and 10 min of centrifugation at 2000 rpm, the (PCA) using a correlation matrix and hierarchical clustering using a supernatant was removed. This procedure was repeated 3 times and the Euclidean similarity index and a paired group (UPGMA) algorithm (Past three extracts were combined. They were then dried and extracted with v4.03). When evaluating posterior distributions, unknown samples were 5 ml of hexane. After 15 min of sonication and 10 min of centrifugation attributed to the type most probably contributing the highest to the at 2000 rpm the supernatants were removed. This procedure was mixture. repeated 3 times. The precipitates (asphaltenes) were analysed by EA- In the second model, the frequently reported Ts/Tm, 29αβ/30αβ and IRMS. Lastly, the compound classes in the hexane soluble fraction 30G/30αβ ratios were used as proxies. The reference values and inter­ were separated into the saturated (first), aromatic (second) and polar pretation guidelines from the first model were maintained. (third) lipid fractions using a silica gel column and a range of solvents In the third model, potential sources were defined using the values of (Bastow et al., 2007). δ13CAsphaltenes, and the Ts/Tm and the G/C30αβH biomarker ratios as The saturated lipidic fraction was then analysed by GC-MS using an proxies. These were chosen because of their widespread availability in Agilent 7890A gas chromatograph coupled to a 5975C mass spectrom­ published studies of modern reference samples. To obtain the source eter. The GC was fitted with a DB-5MS capillary column measuring 30m values required by FRUITS (Table 1), we evaluated 94 modern published length, 250 μm internal diameter and 0.25 μm film thickness. The bitumen samples (Fig. 2) from Central, Southern, North-eastern and injector was operated in splitless mode at 320 ◦ C. The oven temperature North-western Mesopotamia (Connan et al., 1998, 2005, 2006b, 2013, was set at 60 ◦ C for 1 min and then a 4 ◦ C/min temperature ramp 2018; Connan and Nishiaki, 2003) (see supplementary information 3). brought the oven to 320 ◦ C, where it held the temperature for 10 min. The probability that a source contributed more than 50% to the The MS was set in full scan mode acquiring masses between 50 and 600 mixture was calculated by integrating the probability curve between the m/z range. The m/z 191 and m/z 217 ions were used to detect hopanes 50% and 100% percentiles in the posterior distributions produced by the and steranes respectively and relative retention times were used to Bayesian mixing model. Next, a Linear Discriminant Analysis (LDA) was identify the presence of 18α(H)-22,29,30-trisnorhopane (Ts), 17α(H)- applied to these results. This step corrected biases undervaluing or 22,29,30-trisnorhopane (Tm), 17α(H),21β(H)-30-norhopane (29αβ), overvaluing the probability that certain regions were the main source of 18α(H)-oleanane (Oleanane), 17α(H),21β(H)-hopane (30αβ), 17β bitumen in the mixture. The LDA found the region with the shortest (H),21α-(H)-hopane (30βα), 17α(H),21β(H),22S-homohopane (31αβS), Mahalanobis distance to each sample. This distance was calculated using Gammacerane (30G), 24-ethyl-5α(H),14α(H),17α(H),20S-cholestane + a pooled within-group covariance matrix (Hammer et al., 2001) in the 24-ethyl-5α(H),14α(H),17α(H),20R-cholestane (29αα) and 24-ethyl-5α (H),14β(H),17β(H),20S-cholestane + 24-ethyl-5α(H),14β(H),17β (H),20R-cholestane (29ββ) (European Normalisation Comitee, 2009). Table 1 Median and standard deviation values representative of each region. Integrated peak areas were used to calculate the Ts/Tm, Ts/30αβ, Tm/30αβ, 29αβ/30αβ, 30βα/30αβ, 31αβS/30αβ, 30G/30αβ hopane ra­ Region δ13CAsphaltenes Unc. Ts/Tm Unc. G/C30αβH Unc. tios and the 29ββ/29αα sterane ratio. Central − 28.25 0.189 0.130 0.095 0.150 0.080 The isotopic analyses of the asphaltene fractions were undertaken South − 27.30 0.214 0.460 0.102 0.100 0.022 using a Thermo FlashEA 1112 Series elemental analyser coupled to a North-east − 27.60 0.337 0.280 0.230 0.125 0.077 North-west 27.95 1.911 0.955 0.452 0.049 0.146 Thermo Delta V Advantage isotopic ratio mass spectrometer with a − 3 A. Breu et al. Journal of Archaeological Science 145 (2022) 105643 Fig. 2. Modern sources reported in previous studies of ancient bitumen are grouped in four regions: Northwestern Mesopotamia (purple), Norhteastern Mesopotamia (blue), Central Mesopotamia (green) and South (brown). 1. Al Fattha, 2. Abu Jir, 3. Hit, 4. Fallujah, 5. Kilf, 6. Fattah 4, 7. Fattah 1, 8. Fattah 2, 9. Kifri, 10. Mishraq, 11. Qandil Bridge, 12. Hammam al-Aliel, 13. Bai Hassan, 14. Kirkuk 131, 15. Kirkuk 130, 16. Qayyara, 17. Siirt-Eruh 1, 18. Siirt-Eruh 2, 19. Sara Sati, 20. Harbol, 21. Damir Dagh, 22. Kirkuk 117, 23. Kumcati, 24. Zakho, 25. Kerbent, 26. Batman - Gerçüs, 27. Yesilli, 28. Bati Raman (South), 29. Raman, 30. Bogazkoy, 31. Garzan, 32. Magrip, 33. Taq Taq 3, 34. Selmo Oil Seep I, 35. Hazro oil seep, 36. Samsat, 37. Zengen, 38. Selmo II, 39. Neftik, 40. Batman-Yolçati, 41. Adiyaman, 42. Camberlitas, 43. Cukurtas, 44. Kahta, 45. Diyarbakir-Sadi, 46. Diyarbakir-Villa, 47. Diyarbakir-Bismil, 48. Burgan, 49. Barbes, 50. Yenikoy, 51. Kayakoy, 52. Sahaban, 53. Kayakoy-west, 54. Beycan, 55. Kurkan, 56. Sincain, 57. Malatepe, 58. Camurulu, 59. Dincer Guney, 60. Bichri, 61. Aman Hassan, 62. Mamatain, 63. Karun, 64. Haft Kel, 65. Par-E-Siah, 66. Naft Safid, 67. Masjid-I-Suleiman, 68. Lali, 69. Gel-Siah, 70. Chersh Merghir, 71. Shaaf 1, 72. Bichri - Abu Jir, 73. Shaaf 2, 74. Dehluran, 75. Pataq, 76. Dalaki. 77. Sultan Dyke, 78. Sepadhan, 79. Razeh, 80. Ain Gir, 81. Siah Kuh, 82. Dead sea blocks. Inverted triangles: seeps not included in the Bayesian modelling. Sources: Connan and Nishiaki, 2003; Connan et al., 2005, 2013, 2018, 2020, 2022; Hauck et al., 2013. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) PAST v4.03 statistical software. early third millennium and early second millennium was calculated and To test the accuracy of the third model, 57 published modern used to create Figs. 5 and 7. reference samples and 86 previously published archaeological samples were analysed as unknowns. In the first case, 76.8% of the results were 3. Results identical to those published in the literature. In the second case, the rate of success rose to 83.7%. The main sources of error resulted from 3.1. Molecular and isotopic data southern and central Mesopotamian samples mistakenly attributed to north-eastern sources (12.8%) (see supplementary information 5 for The 12 analysed amorphous visible residues contained well-defined more details). Given the high degree of agreement between the refer­ hopane fingerprints and 11 of them also included steranes (Fig. 3) (see ences and the outputs and bearing in mind the limitations of the model, additional chromatograms in supplementary information 2). Their well-preserved bitumen residues from TL were submitted to the same identification confirms the petrogenic origin of the samples, which had analysis. been preliminarily labelled as bitumen based on their black colour and Finally, the intersections of posterior probability distributions were amorphous texture. calculated to quantify the total amount of a bitumen type in a group of Criteria laid out by previous research (Connan et al., 1998, 2005; samples. This, P(A ∩ B) = P(A)⋅P(B), has been successfully applied to Stern et al., 2008; Faraco et al., 2016) assessing the sample’s degrees of both palaeodieary studies (Phillips et al., 2005, 2014; Stock et al., 2018) weathering and biodegradation was used to avoid incorrect geochemical and the analysis of lipid residues in archaeological artefacts (Breu et al., evaluations. The origin of the samples was only considered when the 2021). In this case, the overall quantity of TL’s bitumen types in the C27 to C35 hopanes remained unaltered; steranes could be identified, 4 A. Breu et al. Journal of Archaeological Science 145 (2022) 105643 Fig. 3. Representative hopane (m/z 191) and sterane (m/z 217) distribution (Sample 1). and a low signal to noise ratio was detected. As indicated in Table 2, Additionally, the Ts/Tm, 30βα/30αβ and 30G/30αβ ratios were signifi­ these were samples 1, 5, 6, 7, 8, 9, 10, 11 and 12. A soil sample analysed cantly low while Ts/30αβ was high. to evaluate possible contamination from the environment did not pre­ Type B (sample 8): it included a unique combination of high 29ββ/ sent any detectable traces of hopanes and steranes using the character­ 29αα, Tm/30αβ, Ts/30αβ values with low Ts/Tm, 30G/30αβ, 31αβS/ istic m/z 191 (hopanes) and m/z 217 (steranes) ions 30αβ and 30βα/30αβ values. The 29αβ/30αβ values were placed at the The values of the 8 biomarker ratios obtained from the well- middle of the distribution. preserved samples (Table 3) were consistent with previous analyses of Type C (sample 9): it was characterised by the highest Ts/Tm, 30βα/ archaeological bitumen (Connan and Carter, 2007; Van de Velde et al., 30αβ, 31αβS/30αβ and 30G/30αβ values combined with the lowest Tm/ 2015; Connan et al., 2018). The Ts/Tm ratio spanned between 0.15 and 30αβ ratio. Additionally, the 29ββ/29αα and 29αβ/30αβ ratios were high 0.32 with a median at 0.17; the 29αβ/30αβ ratio presented values be­ while the Ts/30αβ value was low. tween 1.05 and 1.21 with a median in 1.16; the 30G/30αβ ratios were Type D (sample 12): it presented the second highest 29ββ/29αα placed between 0.18 and 0.41 with a median at 0.24, and the δ13C values combined with the lowest 29αβ/30αβ values. Additionally, the isotopic values of the asphaltene fraction ranged within a small range Ts/Tm, 30βα/30αβ, 31αβS/30αβ and 30G/30αβ ratios were high while between − 28.31‰ and − 27.52‰ with − 28.13‰ as its median. Ole­ the Ts/30αβ and Tm/30αβ were low. anane was absent in all the samples. In summary, the hopanes eluting before 30αβ in types D and C pre­ The outcome of the PCA and the groups resulting from the hierar­ sented low ratios while those from compounds eluting afterwards were chical clustering analysis at height = 0.26 suggested the existence of higher. The opposite was observed in the A and B geochemical finger­ four distinct geochemical fingerprints within Tell Lashkir (Fig. 4). prints, and it represented one of the main sources of variation between Furthermore, as differences between these four groups were maximised archaeological bitumen in TL. The Ts/Tm ratio further differentiated the in samples 8, 9, 10 and 12, these were used to define four dominant D and C from the A and B types, and the 29αβ/30αβ hopane and the fingerprint types: 29ββ/29αα sterane ratios were the major sources of variation between D Type A (sample 10): it contained the lowest 29ββ/29αα and 31αβS/ and C, and between A and B. The δ13C isotopic ratios of the asphaltene 30αβ values combined with the highest 29αβ/30αβ and Tm/30αβ values. fractions did not aid in distinguishing between types A, C and D. Its only Table 2 The sample’s archaeological context (SU: stratigraphic unit) and the results of the GC-MS and EA-IRMS analyses with their associated standard deviation. Sample ID Sector SU Hopanes Steranes Oleanane δ13CAsphaltenes (%) Uncertainty 1 6 11.10 Present Present Absent − 27.66 0.03 2 1 3.3 Degraded Degraded Absent Not analysed Not analysed 3 11 44.2 Present Absent Absent Not analysed Not analysed 4 1 31.3 Degraded Degraded Absent Not analysed Not analysed 5 1 31.4 Present Present Absent − 28.17 0.03 6 3 29.1 Present Present Absent − 28.03 0.03 7 3 29.2 Present Present Absent − 28.05 0.03 8 6 11.9 Present Present Absent − 27.52 0.03 9 1 31.3 Present Present Absent − 28.15 0.03 10 1 25.1 Present Present Absent − 28.13 0.03 11 6 11.9 Present Present Absent − 28.31 0.03 12 6 11.9 Present Present Absent − 28.23 0.03 Soil 1 3.3 Absent Absent Absent Not analysed Not analysed 5 A. Breu et al. Journal of Archaeological Science 145 (2022) 105643 significant variation corresponded with type B’s asphaltene δ13C iso­ 0.0200 0.0200 0.0030 0.0300 0.0200 0.0200 0.0200 0.0100 0.0100 topic values being 0.5‰–0.8‰ enriched compared with the other types. Unc. To validate the groupings suggested by the hierarchical clustering analysis and to explore the presence of mixtures, the first Bayesian mixing model incorporated samples 8, 9, 10 and 12 as reference values 29bb/29aa for the A, B, C and D types, respectively. Results indicated that type A 1.2817 1.1529 1.1045 1.2509 1.4352 1.3670 0.9907 1.6096 1.5898 bitumen was the most abundant in the assemblage followed by D, B and C respectively (Fig. 5). Several samples could have resulted from a mixture of two types (Table 4). Sample 1 would have been majorly conformed by type B bitumen with a minor input of type A. Alterna­ 0.0005 0.0030 0.0020 0.0002 0.0020 0.0004 0.0005 0.0020 0.0005 Unc. tively, sample 5 may have contained major amounts of type A and minor inputs of type B. Finally, sample 7 was interpreted as originating from a mixture of equal parts type D and A (Fig. 6). 30G/30 ab 0.2230 0.1851 0.2394 0.2819 0.1889 0.4087 0.2031 0.3739 0.3107 3.2. Exploring trade through published analyses of archaeological bitumen To place TL within the Mesopotamian Bronze Age trade networks, 0.0010 0.0070 0.0050 0.0004 0.0040 0.0008 0.0010 0.0030 0.0010 Unc. the presence of the A, B, C and D geochemical fingerprints was evaluated in 33 published bitumen remains from five archaeological sites located in the 2500-1100 cal. BC time bracket: Kavuşan Höyük (Connan et al., 31abS/30 ab 2013), Salat Höyük (Connan et al., 2018), Umm an Namel, Tell F6 (Connan and Carter 2007) and Ra’S al-Jinz (Connan et al., 2005) (See 0.5257 0.5051 0.5146 0.5917 0.5003 0.7592 0.4944 0.6828 0.6942 supplementary information 6). Contemporary data from Başhur Höyük (Connan et al., 2018) and Tell Abraq (Van de Velde et al., 2017), was not modelled as their studies determined origins (Iran) outside the range of 0.0002 0.0010 0.0008 0.0001 0.0007 0.0002 0.0002 0.0005 0.0002 sources contemplated by the Bayesian mixing model. In this model, the Unc. three hopane ratios more consistently reported in the literature (Ts/Tm, 29αβ/30αβ and 30G/30αβ) were used. Nonetheless, accurate differen­ tiation between types A and B was not clear given their similar Ts/Tm, The archaeological context and the results of the GC-MS and EA-IRMS analyses with their associated standard deviation. 30ba/30 ab 30G/30αβ and 29αβ/30αβ ratios. In the absence of other sources of 0.0906 0.0734 0.0876 0.0783 0.0843 0.1490 0.0740 0.1047 0.1042 discrimination such as the 29ββ/29αα sterane ratio or the Tm/30αβ and the 31αβS/30αβ hopane ratios, these two types were studied as indica­ tive of the same source. The reference values for the A/B group were 0.0040 0.0200 0.0090 0.0010 0.0080 0.0040 0.0060 0.0070 0.0003 thus constructed by averaging samples 8 and 10. Unc. The results of the second Bayesian mixing model indicate that bitumen similar to TL’s type D was only present in sites located around the Persian Gulf. Upriver from TL, the model considered the probability 29ab/30 ab for the presence of this geochemical fingerprint to be below 0.1% in 1.1655 1.1872 1.1487 1.1288 1.1574 1.1620 1.2143 1.0831 1.0514 Bashur Höyük and Salat Höyük and not higher than 2.5% in Kavuşan Höyük. Type C’s fingerprint was not needed by the model to explain the composition of bitumen from sites other than TL. Only Umm an Namel presented a lower than 2.5% chance of it contributing more than 20% to 0.0070 0.0030 0.0008 0.0020 0.0060 0.0001 0.0050 0.0050 0.0030 Unc. a mixture with other sources. Finally, hopane ratios coherent with types A and B were reported in all sites both north and south from TL. Moreover, the model was not able to find a mixture of TL’s types which Tm/30 ab could result in the values obtained either in Bashur Höyük or from 0.2521 0.3193 0.3061 0.1700 0.2886 0.1075 0.3200 0.1565 0.1190 sample 993a in Tell F6. This could indicate that these sites had access to trade routes transporting bitumen types which did not reach TL. Alter­ natively, they may have chemically altered bitumen in a way that made it dissimilar to that from TL. Finally, this types may also be unrepre­ 0.0003 0.0005 0.0007 0.0010 0.0020 0.0010 0.0020 0.0004 0.0008 Unc. sented in our sample collection. 3.3. Exploring the geographic origin of archaeological bitumen Ts/30 ab 0.0431 0.0555 0.0468 0.0326 0.0493 0.0334 0.0493 0.0333 0.0320 Given the absence of oleanane in our samples, it was improbable that TL bitumen originated from Iran or Azerbaijan (Connan and Deschesne, 1995; Connan et al., 1998; Bechtel et al., 2014). Furthermore, an origin 0.0030 0.0002 0.0020 0.0040 0.0040 0.0100 0.0030 0.0040 0.0100 in Oman, where modern references present δ13CAsphaltene values around Unc. − 34‰ (Connan et al., 2005) or Jordan, where δ13CAsphaltene values are placed at − 29‰ (Connan et al., 2006a), would be equally unlikely. Moreover, all reference values from sources in modern Syria and the 0.1708 0.1737 0.1527 0.1915 0.1708 0.3105 0.1541 0.2126 0.2693 TS/Tm Black sea report 30G/30αβ ratios above 0.5 (Connan and Nishiaki, 2003; Hauck et al., 2013; Fulcher et al., 2020), which none of TL’s archaeo­ Table 3 logical samples present. Other sources in Italy (Pennetta et al., 2020) 10 11 12 ID 1 5 6 7 8 9 have not been considered due to their absolute distance from 6 A. Breu et al. Journal of Archaeological Science 145 (2022) 105643 Fig. 4. a) Principal Components Analysis. Dashed ellipses group samples according to b. b) Dendrogram presenting possible groupings coherent with the calculation of the Euclidean similarity index. The dashed line cuts the dendrogram at height 0.26. Table 4 Mean and standard deviation of the quantity of each bitumen type present in samples from TL according with the results of the first Bayesian mixing model. ID A B C D Interpretation 1 28.3 ± 5 58 ± 5 11.5 ± 1 1.7 ± 1 Mainly B, low amounts of A 5 63.5 ± 5 36 ± 5 0.2 ± 0.2 0.5 ± Mainly A, low amounts of 0.5 B 6 83.8 ± 2 2.5 ± 2 12.8 ± 0.8 ± Type A 0.8 0.7 7 46.6 ± 0.1 ± 4±1 49 ± 2 Equal mix between D and 0.7 0.1 A 8 0 100 0 0 Type B 9 0 0 100 0 Type C 10 100 0 0 0 Type A 11 1.5 ± 1 1±1 32 ± 2 65 ± 3 Mainly D, low amounts of C 12 0 0 0 100 Type D north-eastern Mesopotamia. Alternatively, modern sources across southern, central, north- eastern, and north-western Mesopotamia (Fig. 1) presented values resembling those in samples from TL. At this point, the third Bayesian mixing model was used to better distinguish only between the afore­ mentioned zones. The results (Table 5) indicate that bitumen from three different regions might have reached TL. Fig. 5. Boxplot reporting the normalised total quantity of each bitumen type In types A (samples 5, 6, 7, 10) and D (samples 11 and 12), the from the studied assemblage in Tell Lashkir. probability that most of the bitumen in the residue originated from either north-eastern, north-western or southern Mesopotamia was al­ ways below 10%. Alternatively, the likelihood their origin was from 7 A. Breu et al. Journal of Archaeological Science 145 (2022) 105643 Table 5 The probability each region was the major source (>50%) of each bitumen sample from Tell Lashkir according to the results of the third Bayesian mixing model and its interpretation after the LDA analysis. Sample Lashkir Central North South North Interpretation ID type -east -west (after LDA) 1 B+A 0.18 0.49 0.01 0.02 Bitumen from north-eastern Fig. 6. Boxplots present the percentage of bitumen by weight predicted by the first Bayesian mixing model in samples most probably containing a mixture of two types. Mesopotamia 5 A+B 0.68 0.09 0.00 0.01 Bitumen from central Mesopotamia 6 A 0.81 0.08 0.00 0.01 Bitumen from central Mesopotamia 7 D+A 0,70 0.09 0.00 0.01 Bitumen from central Mesopotamia 8 B 0.10 0.42 0.06 0.02 Bitumen from north-eastern Mesopotamia 9 C 0.63 0.02 0.01 0.24 Bitumen from central Mesopotamia 10 A 0.77 0.05 0.01 0.00 Bitumen from central Mesopotamia 11 D+C 0.87 0.03 0.00 0.05 Bitumen from central Mesopotamia 12 D 0.76 0.05 0.00 0.04 Bitumen from central Mesopotamia central Mesopotamia was always higher than 50% and the LDA analysis assigned them to this region. In type B (samples 1 and 8), the highest probability was that most of the bitumen originated from north-eastern Mesopotamia (49% and 42% respectively). The LDA analysis assigned them to this region. Coher­ ently, the model allocated significantly low probabilities to an origin from central, southern, or north-western Mesopotamia. Type C (Sample 9) was placed in a separate category despite the high probability that this bitumen originated from central Mesopotamia (63%), an assumption supported by the LDA analysis. This is because the model did not fully reject the possibility that more than 50% of the residue originated from north-western Mesopotamia (24%). Given that no modern references presented similar biomarker values, it is likely that this sample is the outcome of a mixture of bitumen sourced from these two regions. 4. Discussion 4.1. Strengths and caveats of Bayesian mixing models Across our research into bitumen trade, Bayesian mixing models have provided us the capacity to treat archaeological data as probability distributions reflecting both the analytical results and their associated uncertainties. By assuming all samples are the product of a mixture, the model evaluated the full range of combinations which could explain the obtained values, including the possibility that they not resulted from a mixture in the first place. In our approach, modern oil seeps were grouped in regions which display coherent biomarker and isotopic ra­ tios. Consequently, the Bayesian model would be able to approximate the location of unknown exploited seeps originating from the same crude oil reservoirs as other nearby seeps included in the modern reference set. However, these models may conduce to erroneous interpretations when used as a black box. In our case, the complex composition of petrogenic organic matter implied that at least eight hopane and stear­ ane ratios were necessary to distinguish the four bitumen types present 8 A. Breu et al. Journal of Archaeological Science 145 (2022) 105643 in TL. When this was not possible in the second model, a clear separation compute cases where bitumen originated from a crude oil reservoir between types A and B could not be achieved. Therefore, in cases where missing from the reference dataset. Alternatively, the use of intra-site the model is most prone to error, results should be always carefully and inter-site archaeological references will reflect the structure of evaluated both across modern references and published studies from bitumen trade in Bronze Age Mesopotamia regardless of the exact other archaeological sites. This is the case of samples 1 and 8, tentatively location of each seep, thus avoiding this issue completely. interpreted as originating from north-eastern Mesopotamia despite the Third, the Bayesian mixing models were most effective when model’s bias towards producing false positives from this region. The deployed as ad hoc solutions supported by prior knowledge. A fair un­ archaeological reference samples from southern Mesopotamia which derstanding of the most likely sources and a careful evaluation of the were misidentified by the model as from the North-East had Ts/Tm ra­ model’s sensitivity were key to develop models capable to answer our tios exceeding 0.23 and δ13CAsphaltene values above − 27.5‰. This was archaeological questions. In the case of the third model, reducing the never the case in samples 1 and 8, where the Ts/Tm ratios were lower number of possible sources to four was key to ensure model conver­ and the δ13CAsphaltenes were placed below − 27.5‰. Similarly, all the gence. Therefore, it should not be ruled out that the range of types misidentified archaeological reference samples from central Meso­ detected in TL could be reformulated as additional samples are analysed. potamia, which presented a δ13CAsphaltenes value like that of TL’s Fourth, given that the model outputs are statistical distributions, bitumen, had different Ts/Tm ratios. While we can posit that it is safe to they can be used to test specific archaeological questions. This makes assume that samples 1 and 8 do not constitute false positives, the limi­ possible to claim that the probability that bitumen extracted in the tations of the model did require that this possibility was evaluated. Persian Gulf was a major component in the analysed samples is lower Overall, four major considerations should be taken into account than 6%. This can also be applied to test the similarity of different when applying Bayesian mixing models to the apportionment of samples or groups of samples. In our case, the probability that a sample archaeological bitumen mixtures. First, the model’s discrimination of bitumen from the Bronze Age layers contained major amounts power rises significantly when higher numbers of hopane and sterane (>50%) of the same type than a sample from the MBA layers is 11%. ratios are used as proxies. In our case, the first model was capable of Finally, a posteriori addition of sample results from similar periods or distinguishing two different fingerprints (A and D) which were both archaeological contexts can be used to significantly improve the accu­ assigned to the central Mesopotamian region by the third model. The racy of the results (see error margins in Figs. 6 and 7). second model could not distinguish between types A and B and, addi­ tionally, the uncertainties of the first model posteriors were significantly 4.2. Historic significance of TL’s bitumen remains lower than those of the second and third models. This suggests that there is room for significant improvements in precision and accuracy as more Bearing in mind the relatively low sample size, our results so far hopane and sterane ratios are made available. suggest that bitumen from sectors 3 and 6 might have contained Second, despite bitumen sources mentioned by cuneiform tablets different geochemical fingerprints from sector 1 (Fig. 5). In the Ninevite (Heimpel 2009) and the variability detected in archaeological samples 5 period, type D would have been the most abundant (38–47%), closely are used to infer good coverage of potential past sources by the modern followed by B (23–35%) and A (22–34%). Later, in the MBA, the major reference set, still, cases where archaeological hopane and sterane bio­ type could have shifted to A (43–62%), closely followed by C (33–39%), markers do not match (ex: Connan 1999; Van de Velde 2017b; Connan and the presence of B (1–23%) and D (1–12%) would have been and Oates 2018) point at potential gaps in the reference data. Partially significantly reduced. Nonetheless, Type’s A ubiquity in the three mitigating this problem, Bayesian mixing models offer the opportunity studied sectors suggests that the supply of this source of bitumen could to test whether unmatched samples could be the result of a mixture from have been somewhat stable across time. Major changes in the origin of known seeps, as it was the case with sample 9 in TL, but won’t be able to the bitumen reaching TL would not be unexpected as sectors 3 and 6 Fig. 7. Boxplots present the percentage of bitumen by weight in a) the combination of six samples from the early 3r millennium cal BC b) the combination of three samples from the Middle Bronze Age. 9 A. Breu et al. Journal of Archaeological Science 145 (2022) 105643 predate the formation of the Assyrian imperial core (Ashur-Niniveh- have been combined and recycled when necessary, a practice which was Erbil) while sector 1 would be placed in a period when the first empires not uncommon (Schwartz and Hollander, 2000). had started the integration of territories across Mesopotamia and beyond. 5. Conclusions Neighbouring contemporary archaeological sites including Qalinj Agha (Abu al-Soof, 1966; 1969), Tell Nader (Beuger 2016: 20), Erbil’s The novel implementation of Bayesian mixing models to the study of citadel (Nováček et al., 2008), and later sites such as Kurd Qaburstan archaeological bitumen has been successful in facilitating and (Schwartz et al., 2017), the Darband-i Rania pass (MacGinnis et al., improving the apportionment of mixtures from various sources. The first 2020), and previous landmarks in the region such as Shanidar Cave analyses of bitumen from north-eastern Mesopotamia in the early third (Solecki et al., 2004), Gird Banahilk (Gómez-Bach et al. 2019) or Jarmo millennium (Ninevite 5 horizon) suggest that the population inhabiting (Braidwood and Braidwood 1950) present considerable amounts of Tell Lashkir owned products found in exchange networks which reached bitumen remains. This suggests that, in northern Mesopotamia, bitumen southern Mesopotamia. We propose that Tell Lashkir’s inhabitants had been and was acquired, consumed, and traded by most human traded with bitumen originating from both nearby sources and distant settlements in the region. Resembling the processes in motion in regions including central Mesopotamia. While results support the southern Mesopotamia, this was a product deeply intertwined in the widespread use of central Mesopotamian bitumen in this period, the Bronze age way of life. In TL, the detection of obsidian blades, moulds putative origin of samples 1 and 8 is coherent with written accounts and for metallurgical production and a diverse range of pottery types (Molist geochemical analyses indicating that bitumen from the north-east also et al., 2019a) are coherent with the site’s nature as a small settlement reached southern cities (Connan and Van de Velde, 2010). operating as logistic point through long trade routes crossing Meso­ Bitumen was instrumental to the origin of the first civilizations and potamia. In this sense, TL would not be unique, as other archaeological states, as well as the development of the first global trade networks. As sites including Bardastee (Eidem, 2015), Bash Tapa (Marti et al., 2015) this paper has shown, the use of more refined statistical techniques to or Kani Shaie (Tomé et al., 2015) exhibit similar characteristics (Molist interpret chemical data on archaeological bitumen allows deciphering in et al., 2019a). Nonetheless, TL has offered us the opportunity to probe exceptional detail the apportionment of ancient bitumen trade in south- the range of bitumen types in circulation across the region. Results from western Asia. Thus, the combined biomarker and Bayesian approach our second mixing model indicate that TL’s occupants were involved in offers new insights into the origin of civilisation and the first states, one Mesopotamian-wide bitumen trade networks, as exemplified by types A of the defining periods in the history of humankind. and B. In the early third millennium, they were able to acquire products which also reached the Persian Gulf (type D bitumen) but were not Author contributions present upriver (Kavuşan Höyük, Salat Höyük and Başur Höyük, Connan et al., 2013; 2018). Even though the sample size is still low, the absence AB, AGB, ARM and MM designed the project and funded the of oleanone, and hopane, sterane and asphaltene values from modern research; AB and ARM planned the research, performed the biomarker Azerbaijan or Iran should not be underestimated. This set of results analysis, and processed the data; AGB and MM provided samples and foreshadows the tighter relationship this region would develop with contextual data; AB wrote the manuscript with contributions from all southern Mesopotamia when, in the third millennium, the first written other co-authors. record of the city of Erbil appeared in an inscription from Nippur (MacGinnis 2014). Declaration of competing interest During the Early Dinastic Period, tablets from Girshu, Umma and Gudea indicate that expeditions consisting of up to one hundred men The authors declare that they have no known competing financial were sent to obtain copious quantities of bitumen upriver to either interests or personal relationships that could have appeared to influence northern or central Mesopotamian seeps (Heimpel 2009). The analysis the work reported in this paper. of bitumen remains from Susa suggests that, across time, the kingdom of Elam acquired bitumen from closer sources in the Susiana plain (Connan Acknowledgments and Deschesne 1996). At TL, the outcomes of the third model indicate that central Mesopotamian seeps (types A and D) might have been Carl Heron at The British Museum is thanked for his support in the contributing bitumen to the Tigris trade networks significantly more identification of the visible pottery residues. We also thank the Sala­ than their north-eastern Mesopotamian counterparts. Nonetheless, re­ haddin University and the Autonomous University of Barcelona exca­ sults from samples 1 and 8 suggest that settlements in the upper Tigris vation teams in Tell Lashkir for their hard work. Adrià Breu has been might still have played a significant role as producers of this raw ma­ funded by an FPU grant (2014/00957). Analyses were partially funded terial. In the Euphrates, analyses from Tell Brak (Connan and Oates by the following I + D projects awarded by the Spanish Ministry of 2018) and Mari (Connan and Deschesne 2007) suggest bitumen was also Education and Science (PID2019-106399 GB-I00, HAR2016-78416-P acquired from a combination of northern and central Mesopotamian and HAR2013-43624-P) and the T002019N0000038379 project awar­ seeps. ded by the Instituto de Patrimonio Cultural de España. Anna Gómez- It is still difficult to assess the specific role bitumen played in TL. So Bach, Miquel Molist and Adrià Breu are members of the 2017 SGR 1302 far, their absence in architecture and their overwhelming presence as a Research Group recognized by the Generalitat de Catalunya. Pau Comes, coating in pottery suggests it might have been commonly used as an Nuria Moraleda and Ferran Colomer are acknowledged for the EA-IRMS adhesive or as a waterproofing agent in pottery. The absence of petro­ analyses and support in the GC-MS analyses at the ICTA-UAB graphic studies prevents assessing whether the provenance of its ceramic laboratories. containers could be tied with the trade of this product. Nonetheless, it is our interpretation that, given that the Bronze age buildings in sector 1 Appendix A. Supplementary data are coherent with storage spaces, it should not be ruled out that part of the detected bitumen was not used, but stored in large jars waiting for Supplementary data to this article can be found online at https://doi. the right trading conditions. Rather than importing significant quanti­ org/10.1016/j.jas.2022.105643. ties of bitumen from a single source, the Bronze Age inhabitants of TL acquired it from several regions, a strategy which could have mitigated References the risk of losing access to this raw material. Moreover, the possible detection of mixtures (samples 1, 5 and 7) suggests that bitumen could Abu al-Soof, B., 1966. Short soundings at Tell Qalinj Agha (Erbil). Sumer 22, 77–82. 10 A. Breu et al. Journal of Archaeological Science 145 (2022) 105643 Abu al-Soof, B., 1969. Excavations at Tell Qalinj Agha (Erbil). Sumer 25, 3–42. Hartz, S., Kabaciński, J., Kotula, A., Kukawka, S., Loze, I., Mazurkevich, A., Bastow, T.P., Aarssen, B.G.K., van, Lang, D., 2007. Rapid small-scale separation of Piezonka, H., Piličiauskas, G., Sørensen, S.A., Talbot, H.M., Tkachou, A., saturate, aromatic and polar components in petroleum. Org. Geochem. 38, Tkachova, M., Wawrusiewicz, A., Meadows, J., Heron, C.P., Craig, O.E., 2020. 1235–1250. https://doi.org/10.1016/J.ORGGEOCHEM.2007.03.004. Organic residue analysis shows sub-regional patterns in the use of pottery by Bechtel, A., Movsumova, U., Pross, J., Gratzer, R., Ćorić, S., Sachsenhofer, R.F., 2014. Northern European hunter–gatherers. R. Soc. Open Sci. 7, 192016 https://doi.org/ The Oligocene Maikop series of Lahich (eastern Azerbaijan): paleoenvironment and 10.1098/rsos.192016. oil–source rock correlation. Org. Geochem. 71, 43–59. https://doi.org/10.1016/j. Eidem, J., 2015. Bardastee (NINO Archaeological Project on the Rania Plain). orggeochem.2014.04.005. Archaeological Projects in the Kurdistan Region in Iraq. Directorate of Antiquities of Beuger, C., 2016. Short notes on chalcolithic pottery research: the pottery sequences of the Kurdistan Regional Government., p. 9 Tell nader (Erbil) and ashur (Qal’at Sherqat). In: Kopanias, K., Macginnis, J. (Eds.), Faraco, M., Pennetta, A., Fico, D., Eramo, G., Beqiraj, E., Muntoni, M., De Benedetto, G. The Archaeology of the Kurdistan Region of Iraq and Adjacent Regions. E., 2016. Bitumen in potsherds from two Apulian Bronze Age settlements, Monopoli Archaeopress, Oxford, pp. 19–27. and Torre Santa Sabina: Composition and origin. Organic Geochemistry. https://doi. Braidwood, R.J., Braidwood, L., 1950. Jarmo: A Village Early Farmers in Iraq. Antiquity org/10.1016/j.orggeochem.2016.01.004. 24, 189–195. https://doi.org/10.1017/S0003598X00023371. Fernandes, R., Eley, Y., Brabec, M., Lucquin, A., Millard, A., Craig, O.E., 2018. Breu, A., Alcántara, R., Arnaiz, R., Colomer, F., Moya, A., Sisa, J., 2019. Detection, Reconstruction of prehistoric pottery use from fatty acid carbon isotope signatures recovery and analysis of visible Bitumen residues in pottery from Tell Lashkir, using Bayesian inference. Org. Geochem. 117, 31–42. https://doi.org/10.1016/j. Kurdistan, Iraq. In: Álvarez, A., Bello, P., Francés, M., Herrejón, Á., Martínez, M., orggeochem.2017.11.014. Pedro, G. de, Pereira, A., Prado, O., Ruiz, A., Sánchez, Z., Varela, Á., Vidal, M., Fernandes, R., Grootes, P., Nadeau, M.-J., Nehlich, O., 2015. Quantitative diet Zorrilla, G. (Eds.), Los Tiempos Cambian, de La Piedra Al Teclado. X Jornadas de reconstruction of a Neolithic population using a Bayesian mixing model (FRUITS): Jóvenes En Investigación Arqueológica Burgos 7-10 Junio 2017. Asociación de The case study of Ostorf (Germany). Am. J. Phys. Anthropol. 158, 325–340. https:// Historia y Arqueología de Burgos, Burgos, pp. 566–577. doi.org/10.1002/ajpa.22788. Breu, A., Gómez-Bach, A., Heron, C., Rosell-Melé, A., Molist, M., 2021. Variation in Forbes, R.J., 1964. Studies in Ancient Technology. Brill, Leiden. pottery use across the Early Neolithic in the Barcelona plain. Archaeol. Anthropol. Fulcher, K., Stacey, R., Spencer, N., 2020. Bitumen from the Dead Sea in Early Iron Age Sci. 13, 53. https://doi.org/10.1007/s12520-021-01296-2. Nubia. Scientific Reports 10 (1), 1–12. https://doi.org/10.1038/s41598-020-64209- Connan, J., 1999. Use and trade of bitumen in antiquity and prehistory: molecular 8. archaeology reveals secrets of past civilizations. Phil. Trans. Biol. Sci. 354, 33–50. Gómez-Bach, A., Bradosty, Z., Zebari, A.M., Ahmed, H., Cruells, W., Sisa, J., Moya, A., https://doi.org/10.1098/rstb.1999.0358. Breu, A., Molist, M., 2018. New archaeological project: Gird Lashkir: Preliminary Connan, J., Carter, R., 2007. A geochemical study of bituminous mixtures from Failaka data related to Ninevite V period: 2016 and 2017 Fieldwork excavations. In: and Umm an-Namel (Kuwait), from the early Dilmun to the early Islamic period. Bradosty, Z., Zamua, D. (Eds.), Proceedings of the 2nd International Scientific Arabian Archaeol. Epigr. 18, 139–181. https://doi.org/10.1111/j.1600- Conference: Archaeology and Heritage of Hawler-Erbil. Salahaddin University, Erbil, 0471.2007.00283.x. Iraq, pp. 331–345. Connan, J., Deschesne, O., 1995. Archaeometric study of bituminous artefacts of Susa Gómez-Bach, A., Breu, A., Sisa, J., Mylona, P., Bradosty, Z., Zebari, A., Molist, M., 2019a. (Iran). In: Grimalt, J., Dorronsoro, C. (Eds.), Organic Geochemistry: Developments Singular deposits in Gird Lashkir (Erbil, Iraqui Kurdistan) at the beginning of the and Applications to Energy, Climate, Environment and Human History. European Early Bronze Age. In: 3rd International Scientific Conference: Archaeology and Association of Organic Geochemists, Donostia, pp. 708–711. Heritage of Kurdistan, pp. 551–561. Connan, J., Deschesne, O., 1996. Le Bitume a Suse. Collection du musée du Louvre. Elf- Gómez-Bach, A., Cruells, W., Alcàntara, R., Saña, M., Douché, C., Molist, M., 2019b. New Aquitanie, Publications, Paris. excavations at Gird Banahilk, a Halafian site in Iraqi Kurdistan: Farmer and herder Connan, J., Deschesne, O., 2007. Le bitume à Mari. communities in the Upper Zagros Mountains. Paleorient 45.2, 53–66. Connan, J., Nishiaki, Y., 2003. The bituminous mixtures of Tell Kosak Shamali on the Hammer, Ø., Harper, D.A.T., Ryan, P.D., 2001. PAST: Paleontological statistics software Upper euphrates (Syria) from the Early Ubaid to the Post Ubaid: composition of package for education and data analysis. Palaeontol. Electron. 4, 9. mixtures and origin of bitumen. In: Nishiaki, Y., Matsutani, R. (Eds.), The Hauck, T.C., Connan, J., Charrié-Duhaut, A., Tensorer, J.M., Sakhel, Le, Al, H., 2013. Archaeological Investigations on the Upper Euphrates, Syria, Chalcolithic Molecular evidence of bitumen in the Mousterian lithic assemblage of Hummal Technology and Subsistence. The University Museum and the University of Tokyo., (Central Syria). J. Archaeol. Sci. 40, 3252–3262. https://doi.org/10.1016/j. Tokyo, Japan, pp. 283–306. jas.2013.03.022. Connan, J., Oates, J., 2018. The bitumen of Tell Brak from the Middle Uruk (c.3500 BC) Heimpel, W., 2009. The Location of Madga. J. Cuneif. Stud. 61, 25–61. to Late Bronze Age (c.1280 BC): origin and trade routes. J. Hist. Archaeol. Lucquin, A., Robson, H.K., Eley, Y., Shoda, S., Veltcheva, D., Gibbs, K., Heron, C.P., Anthropol. Sci. 3 https://doi.org/10.15406/jhaas.2018.03.00163. Isaksson, S., Nishida, Y., Taniguchi, Y., Nakajima, S., Kobayashi, K., Jordan, P., Connan, J., Lombard, P., Killick, R., Højlund, F., Salles, J.-F., Khalaf, A., 1998. The Kaner, S., Craig, O.E., 2018. The impact of environmental change on the use of early archaeological bitumens of Bahrain from the Early Dilmun period (c.2200 BC) to the pottery by East Asian hunter-gatherers. Proc. Natl. Acad. Sci. USA. https://doi.org/ sixteenth century AD: a problem of sources and trade. Arabian Archaeol. Epigr. 9, 10.1073/pnas.1803782115, 201803782. 141–181. https://doi.org/10.1111/j.1600-0471.1998.tb00116.x. MacGinnis, J., 2014. A City from the Dawn of History: Erbil in the Cuneiform Sources. Connan, J., Carter, R., Crawford, H., Tobey, M., Charrié-Duhaut, A., Jarvie, D., Oxbow Books. Albrecht, P., Norman, K., 2005. A comparative geochemical study of bituminous MacGinnis, J., Raheem, K.R., Ismael, B.B., Ahmad, M., Cabral, R., Dusting, A., boat remains from H3, As-Sabiyah (Kuwait), and RJ-2, Ra’s al-Jinz (Oman). Arabian Greenfield, T., Hazell, G., Iasonos, A., Kertai, D., Miller, A., Pabeschitz, V., Archaeol. Epigr. 16, 21–66. https://doi.org/10.1111/j.1600-0471.2005.00041.x. Shepperson, M., Cartwright, C., Čibera, J., Curtis, V., Dyer, J., Koek, E., Giannese, A., Connan, J., Nissenbaum, A., Imbus, K., Zumberge, J., Macko, S., 2006a. Asphalt in iron Higgs, P., Yusuf, A., Morton, K., Pereira-Pardo, L., Proctor, L., Williams, C., 2020. age excavations from the Philistine Tel Miqne-Ekron city (Israel): Origin and trade Excavations At the Darband-I Rania Pass, Kurdistan Region of Iraq: Report on the routes. Org. Geochem. 37, 1768–1786. https://doi.org/10.1016/j. 2016 and 2017 Seasons. Iraq 82, 139–178. https://doi.org/10.1017/irq.2019.11. orggeochem.2006.08.015. Marti, L., Nicolle, C., Shawaly, K., 2015. Recherches en Haute-Mésopotamie II. Mission Connan, J., Genç, E., Orhan, K., Engel, M.H., Zumberge, A., 2022. Geochemistry and archéologique de Bash Tapa (campagnes 2012-2013) et les enjeux de la recherche origin of bituminous samples of Kuriki H ¨oyük (SE Turkey) from 4000 BCE to 200 dans la région d’Erbil.. SEPOA. CE: Comparison with Kavus ̧ an H ¨oyük, Hakemi Use and Salat Tepe. Journal of Molist, M., 2015. Tell Lashkry: Late Prehistory in the Erbil Region. In: Kopanias, K., Archaeological Science: Reports 41, 103348. https://doi.org/10.1016/j. Macginnis, J., Ur, J. (Eds.), Archaeological Projects in the Kurdistan Region in Iraq. jasrep.2022.103348. Directorate of Antiquities of the Kurdistan Regional Government., Erbil, p. 28. Connan, J., Kavak, O., Akin, E., Yalçin, M.N., Imbus, K., Zumberge, J., 2006b. Molist, M., Bradosty, Z., Breu, A., Sisa, J., Alcántara, R., Cruells, W., Douché, C., Identification and origin of bitumen in Neolithic artefacts from Demirköy Höyök Mylona, P., Arnaiz, R., Saña, M., Zebari, A., Gómez-Bach, A., 2019a. New data on the (8100 BC): Comparison with oil seeps and crude oils from southeastern Turkey. Org. 4th - 3rd millennia in Northern Mesopotamia : The ancient occupations at Gird Geochem. 37, 1752–1767. https://doi.org/10.1016/j.orggeochem.2006.07.023. Lashkir in their archaeological contexts. Paleorient 45 (2), 191–206. Connan, J., Kozbe, G., Kavak, O., Zumberge, J., Imbus, K., 2013. The bituminous Molist, M., Gómez Bach, A., Bradosty, Z., Anfruns, J., Cruells, W., Alcántara, R., mixtures of Kavuşan Höyük (SE Turkey) from the end of the 3rd millennium Arnaiz, R., Breu, A., Moya, A., Sisa, J., Aziz, M., Ahmed, H., 2019b. Proyecto (2000BC) to the Medieval period (AD 14th century): Composition and origin. Org. arqueológico en Gird Lashkir (Erbil, Kurdistan Iraq): resultados de las campañas de Geochem. 54, 2–18. https://doi.org/10.1016/j.orggeochem.2012.09.007. excavación 2015-2016. Inf. y Trab. 17, 201–221. Connan, J., Kavak, O., Sağlamtimur, H., Engel, M., Zumberge, A., Zumberge, J., 2018. Moorey, P.R.S., 1994. Bitumen in building and other roles. In: Ancient Mesopotamian A geochemical study of bitumen residues on ceramics excavated from Early Bronze Materials and Industries: the Archaeological Evidence. Eisenbrauns, Winona Lake, age graves (3000–2900 BCE) at Başur Höyük in SE Turkey. Org. Geochem. 115, United States of America, pp. 332–335. 1–11. https://doi.org/10.1016/j.orggeochem.2017.09.007. Normalisation Comitee, European, 2009. Oil spill identification — Waterborne Connan, J., Priestman, S., Vosmer, T., Komoot, A., Tofighian, H., Ghorbani, B., Engel, M. petroleum and petroleum products — Part 2: Analytical methodology and H., Zumberge, A., Van de Velde, T., 2020. Geochemical analysis of bitumen from interpretation of results based on GC-FID and GC-MS low resolution analyses. In: PD West Asian torpedo jars from the c. 8th century Phanom-Surin shipwreck in CEN/TR 15522-2:2012. BSI Standards Publication, p. 136. Thailand. Journal of Archaeological Science 117, 105111. https://doi.org/10.1016/ Nováček, K., Chabr, T., Filipskỳ, D., Janíček, L., Pavelka, K., Šída, P., Trefnỳ, M., j.jas.2020.105111. Vařeka, P., 2008. Research of the Arbil Citadel, Iraqi Kurdistan, First Season. Connan, J., Van de Velde, T., 2010. An overview of bitumen trade in the Near East from Památky Archeol 99, 259–302. the Neolithic (c.8000 BC) to the early Islamic period. Arabian Archaeol. Epigr. 21, Phillips, D.L., Newsome, S.D., Gregg, J.W., 2005. Combining sources in stable isotope 1–19. https://doi.org/10.1111/j.1600-0471.2009.00321.x. mixing models: Alternative methods. Oecologia 144, 520–527. https://doi.org/ Courel, B., Robson, H.K., Lucquin, A., Dolbunova, E., Oras, E., Adamczak, K., 10.1007/s00442-004-1816-8. Andersen, S.H., Astrup, P.M., Charniauski, M., Czekaj-Zastawny, A., Ezepenko, I., 11 A. Breu et al. Journal of Archaeological Science 145 (2022) 105643 Phillips, D.L., Inger, R., Bearhop, S., Jackson, A.L., Moore, J.W., Parnell, A.C., Stock, B.C., Jackson, A.L., Ward, E.J., Parnell, A.C., Phillips, D.L., Semmens, B.X., 2018. Semmens, B.X., Ward, E.J., 2014. Best practices for use of stable isotope mixing Analyzing mixing systems using a new generation of Bayesian tracer mixing models. models in food-web studies. Can. J. Zool. 92, 823–835. https://doi.org/10.1139/cjz- PeerJ 6, e5096. https://doi.org/10.7717/peerj.5096. 2014-0127. Stol, M., 2012. Bitumen in ancient mesopotamia. The textual evidence. Bibl. Orient. Schwartz, M., Hollander, D., 2000. Annealing, Distilling, Reheating and Recycling : 48–60. Bitumen Processing in the Ancient Near East. Paleorient 26, 83–91. Tomé, A., Cabral, R., Renette, S., 2015. Kani Shaie archaeological excavations. In: Schwartz, M., Hollander, D., 2016. The Uruk expansion as dynamic process: A Kopanias, K., MacGinnis, J., Ur, J. (Eds.), Archaeological Projects in the Kurdistan reconstruction of Middle to Late Uruk exchange patterns from bulk stable isotope Region in Iraq. Directorate of Antiquities of the Kurdistan Regional Government., analyses of bitumen artifacts. J. Archaeol. Sci.: Report 7, 884–899. https://doi.org/ Erbil, p. 24. 10.1016/j.jasrep.2016.01.027. Van de Velde, T., 2017. Provenance Studies on the Bitumen from Tol-e Ajori. In: Schwartz, G.M., Brinker, C.D., Creekmore, A.T., Feldman, M.H., Smith, A., Weber, J.A., Askari, A., Callieri, P. (Eds.), From Palace to Town. Report on the Multidisciplinary 2017. Excavations at Kurd Qaburstan, a second millennium B.C. urban site on the Project Carried Out by the Iranian-Italian Joint Archaeological Mission on the Erbil plain. Iraq 79, 213–255. https://doi.org/10.1017/irq.2017.2. Persepolis Terrace (Fars, Iran), 2008-2013. BraDypUS, Roma, Italy, pp. 45–51. Solecki, R.S., Solecki, R.L., Agelarakis, A.P., 2004. The Proto-Neolithic Cemetery in Van de Velde, T., De Vrieze, M., Surmont, P., Bodé, S., Drechsler, P., 2015. A geochemical Shanidar Cave. Texas A & M University anthropology series. Texas A & M University study on the bitumen from Dosariyah (Saudi-Arabia): Tracking Neolithic-period Press. bitumen in the Persian Gulf. J. Archaeol. Sci. 57, 248–256. https://doi.org/10.1016/ Stern, B., Connan, J., Blackelock, E., Jackman, R., Coningham,, R.A.E., Heron, C., 2008. j.jas.2015.03.007. From Susa to Anuradhapura: Reconstructing aspects of trade and exchange in Van de Velde, T., Magee, P., Lynen, F., 2017. The bitumen imports at Tell Abraq - tracing bitumen-coated ceramic vessels between Iran and Sri Lanka from the third to the the second-millennium BC bitumen industry in south-east Arabia. In: Proceedings of ninth centuries AD. Archaeometry 50 (3), 409–428. https://doi.org/10.1111/j.1475- the Seminar of Arabian Studies. Archaeopress, London, pp. 227–234. 4754.2007.00347.x. 12