geosciences Article A Predictive Model for Maceral Discrimination by Means of Raman Spectra on Dispersed Organic Matter: A Case Study from the Carpathian Fold-and-Thrust Belt (Ukraine) Andrea Schito 1,2,* , Alexandra Guedes 3 , Bruno Valentim 3 , Amanda Vergara Sassarini 1 and Sveva Corrado 1 1 Dipartimento di Scienze, Sezione Scienze Geologiche, Università degli Studi di Roma Tre, Largo San Leonardo Murialdo, 1, 00146 Rome, Italy; [email protected] (A.V.S.); [email protected] (S.C.) 2 Department of Geology and Petroleum Geology, School of Geosciences, University of Aberdeen, Aberdeen AB24 3UE, UK 3 Insituto da Ciências da Terra e Departamento de Geosciências, Ambiente e Ordenamento do Territòrio, Faculdade de Ciências, Universidade do Porto, 4169-007 Porto, Portugal; [email protected] (A.G.); [email protected] (B.V.) * Correspondence: [email protected] Abstract: In this study, we propose a predictive model for maceral discrimination based on Raman spectroscopic analyses of dispersed organic matter. Raman micro-spectroscopy was coupled with optical and Rock-Eval pyrolysis analyses on a set of seven samples collected from Mesozoic and Cenozoic successions of the Outer sector of the Carpathian fold and thrust belt. Organic petrography and Rock-Eval pyrolysis evidence a type II/III kerogen with complex organofacies composed by the Citation: Schito, A.; Guedes, A.; coal maceral groups of the vitrinite, inertinite, and liptinite, while thermal maturity lies at the onset Valentim, B.; Vergara Sassarini, A.; of the oil window spanning between 0.42 and 0.61 Ro%. Micro-Raman analyses were performed, on Corrado, S. A Predictive Model for approximately 30–100 spectra per sample but only for relatively few fragments was it possible to Maceral Discrimination by Means of Raman Spectra on Dispersed Organic perform an optical classification according to their macerals group. A multivariate statistical analysis Matter: A Case Study from the of the identified vitrinite and inertinite spectra allows to define the variability of the organofacies and Carpathian Fold-and-Thrust Belt develop a predictive PLS-DA model for the identification of vitrinite from Raman spectra. Following (Ukraine). Geosciences 2021, 11, 213. the first attempts made in the last years, this work outlines how machine learning techniques have https://doi.org/10.3390/ become a useful support for classical petrography analyses in thermal maturity assessment. geosciences11050213 Keywords: Raman spectroscopy; dispersed organic matter; vitrinite reflectance; principal component Academic Editors: Ian Coulson and analysis; partial least square discriminant analysis; machine learning Jesus Martinez-Frias Received: 24 March 2021 Accepted: 7 May 2021 1. Introduction Published: 14 May 2021 The analysis of coals and dispersed organic matter (DOM) is an irreplaceable tool in Publisher’s Note: MDPI stays neutral thermal maturity assessment and environmental studies of sedimentary successions [1]. In with regard to jurisdictional claims in thermal model calibration for petroleum exploration and basin analysis, vitrinite reflectance published maps and institutional affil- is still the most accurate and reliable indicator [2–5]. On the other hand, vitrinite discrimina- iations. tion among different organic facies has always been a critical issue for a correct assessment of thermal maturity [6,7], in particular when other low reflecting macerals occur [8–11]. Given this and the occurrence of other pitfalls that can interfere with vitrinite reflectance assessment (e.g., suppression and retardation [12,13] and references therein), alternative methods have been proposed, such as palynomorph darkness index (PDI, [14,15]), FT-IR Copyright: © 2021 by the authors. Licensee MDPI, Basel, Switzerland. spectroscopy [16–20], NMR [21], XPS [22], or Raman spectroscopy [23–33]. This article is an open access article In the last years, an increasing interest in the application of Raman spectroscopy for distributed under the terms and thermal maturity assessment in catagenesis and metagenesis [24,26,27,29,31,34–38] has conditions of the Creative Commons been observed. The majority of articles focused on the analysis of single macerals, mainly Attribution (CC BY) license (https:// vitrinite in coals [24,26,27], whereas other ones on bulk kerogen [29,34,38]. Nevertheless, creativecommons.org/licenses/by/ only a few authors dealt with the heterogeneity that can characterize organic facies [23,32] 4.0/). or even a single organoclast [37,39]. Geosciences 2021, 11, 213. https://doi.org/10.3390/geosciences11050213 https://www.mdpi.com/journal/geosciences Geosciences 2021, 11, x FOR PEER REVIEW 2 of 18 been observed. The majority of articles focused on the analysis of single macerals, mainly vitrinite in coals [24,26,27], whereas other ones on bulk kerogen [29,34,38]. Nevertheless, only a few authors dealt with the heterogeneity that can characterize organic facies [23,32] or even a single organoclast [37,39]. Geosciences 2021, 11, 213 Most of the works dealing with Raman spectra of kerogen use a classic band2 offitting 19 approach that in many cases can result in biased results, as recently outlined [36,40]. Thus, different automatic approaches have been proposed based on simplified band deconvolutionMost of the [31,38,41] works dealing or on a withmultivariate Raman spectra analysis of chemometric kerogen use a scheme classic band[34,36] fitting for the spectraapproach analysis, that in proposing many cases a can challenging result in biased approach results, for asgeological recently outlined studies. [In36 ,particular,40]. Thus, a principaldifferent automaticcomponent approaches analysis have(PCA)–partial been proposed least based square on simplified(PLS)-based band multivariate deconvo- analysislution [ 31has,38 been,41] or recently on a multivariate proposed analysis[34,36] to chemometric correlate predictive scheme [34 parameters,36] for the spectrafrom PLS regressionanalysis, proposingagainst vitrinite a challenging reflectance. approach More forover, geological Schito et studies. al. [32] In demonstrated particular, a prin- how a partialcipal componentleast square analysis discrimination (PCA)–partial analysis least (PLS-DA) square (PLS)-based allows to multivariatedefine and predict analysis the differenceshas been recently between proposed vitrinite [34 and,36 ]sporomorphs to correlate predictive on the base parameters of their Raman from PLS spectrum regression for a givenagainst thermal vitrinite maturity reflectance. degree. Moreover, Schito et al. [32] demonstrated how a partial least squareIn this discrimination work, we test analysis a simi (PLS-DA)lar approach allows on to a define set of andseven predict samples the differences collected in be- the tween vitrinite and sporomorphs on the base of their Raman spectrum for a given thermal Outer sector of the Carpathian Fold and Thrust belt (Ukraine), in order to define maturity degree. quantitatively the differences in Raman spectra between vitrinite and inertinite group In this work, we test a similar approach on a set of seven samples collected in the Outer maceralssector of theand Carpathian predict them Fold with and Thrust the aim belt to (Ukraine), provide ina ordermore torobust define dataset quantitatively for thermal the maturitydifferences assessment. in Raman spectra between vitrinite and inertinite group macerals and predict them with the aim to provide a more robust dataset for thermal maturity assessment. 2. Materials and Methods 2.1.2. MaterialsMaterials and Methods 2.1. Materials Samples were collected from black shales at various stratigraphic intervals (Early CretaceousSamples to Early were Miocene) collected cropping from black out shales in the at Outer various Carpathians, stratigraphic in SW intervals Ukraine, (Early close Cretaceous to Early Miocene) cropping out in the Outer Carpathians, in SW Ukraine, to the Romania border (Figure 1). This portion of the fold-and-thrust belt shows an close to the Romania border (Figure1). This portion of the fold-and-thrust belt shows an imbricate fan architecture made up of a series of NE-verging thrust sheets [42] in which imbricate fan architecture made up of a series of NE-verging thrust sheets [42] in which threethree tectonic tectonic units units are are recognized (Boryslav-Pokuttia(Boryslav-Pokuttia tectonic tectonic Unit, Unit, Skiba Skiba tectonic tectonic Unit Unit andand Chornogora Chornogora tectonic tectonic Unit, Unit, Fi Figuregure1 1).). InIn detail, detail, five five samples samples (PL (PL 93.1, 93.1, PL PL 93.2, 93.2, PL PL 95, 95, PLPL 97, 97, PL PL 101.1) 101.1) come come from from organic-rich levelslevels of of the the Melinite Melinite shales shales (Oligocene (Oligocene to Lowerto Lower Miocene)Miocene) cropping cropping out inin the the Skiba Skiba Unit. Unit. Sample Sample PL PL 102 102 was was collected collected in a pelitic in a pelitic bed of bed the of theEocene-Oligocene Eocene-Oligocene succession succession of the of Globigerina the Globigerina Marls inMarls the Skiba in the Unit, Skiba whereas Unit, sample whereas samplePL 103 PL comes 103 fromcomes an from organic-rich an organic-rich level of the level Hauterivian-Albian of the Hauterivian-Albian Spas-Shypot Spas-Shypot formation formationin the Chornogora in the Chornogora Unit (Figure Unit1). (Figure 1). FigureFigure 1. 1.GeographicGeographic and and geologic geologic setting setting of of the the studied studied area: (aa)) SimplifiedSimplified tectonic tectonic sketch sketch of of the