Carbonate systems are influenced by a great variety of physical and biological controlling factors that operate from global to local scales. The resulting intrinsic complexity of carbonate platforms makes them difficult to predict, especially when data are limited. Predicting geologic geometries and properties based on limited sampling or uncalibrated seismic data generally relies on a priori knowledge and equivocal interpretations that are marked by geologist perception and personal experience. To overcome these uncertain interpretations of such a complex natural system, which can become critical in frontier exploration, we developed an expert system that relies on a process-based method and a standardized data set using normalized information and parameters. The main innovation relies on the realization of knowledge- and process-based synthetic carbonate stratigraphic architectures that support seismo-stratigraphic interpretations. The workflow consists of four steps: (1) bibliographic compilation of a geologic database for each case study supported by quantitative parameters (e.g., sedimentation duration and thickness) and qualitative parameters (geodynamic context, seismic architecture, and facies model); (2) statistical analyses to establish consistent geologic classes and spatiotemporal trends; (3) process-based modeling to simulate stratigraphic architectures associated with carbonate sedimentation processes in a physically constrained numerical environment and testing different geologic hypotheses; and (4) realization of a predictive palaeogeographic map representing the global distribution of carbonate stratigraphic architectures, and estimation of controlling parameters for unconstrained case studies. The expert system is based on 77 case studies of Upper Jurassic carbonate platforms, which reveal the resemblance of these carbonate systems, in response to uniform global palaeoclimatic conditions and sea level. Significant local differences in stratigraphic architectures are related to specific geodynamic contexts and subsidence trends. The thickest carbonate platforms are developed in extensive/passive geodynamic settings such as the Central Atlantic Ocean margins, while thinner platforms form in intra- and peri-cratonic settings such as those of the Arabian region.

In exploration, and particularly frontier exploration, the prediction and characterization of potential carbonate reservoir prospects are generally based on seismic interpretation and a priori regional knowledge. The complex relationships between carbonate sedimentary facies and geophysical properties (Paterson et al., 2008; Cantrell et al., 2015; Berra et al., 2016) explain the significant uncertainty of seismic inversion and interpretation of carbonate sequences (Burgess et al., 2006; Warrlich et al., 2008; Montaggioni et al., 2015). Seismo-stratigraphic methods are usually applied to interpret possible stratigraphic architectures and carbonate sedimentary systems (e.g., platform, reefs, or mounds) (Shuster and Aigner, 1994; Eberli et al., 2004; Cacas et al., 2008). In this challenging context, information about the paleogeography and the sedimentology of the interpreted geologic period is key to constraining seismic interpretation (Figure 1). Hence, global-scale studies represent precious analyses to better understand and predict carbonate systems. Such studies dealing with spatial and temporal trends of carbonate platforms throughout the Phanerozoic were conducted through industrial and academic projects (e.g., Markello et al., 2008; Davies et al., 2019; Michel et al., 2019; Kiessling and Krause, 2022). These studies help define deterministic models on which the prediction of carbonate reservoirs is founded (Wilson, 1975; Schlager, 2000; Michel et al., 2020). For instance, global spatial trends of stratigraphic architectures of carbonate platforms, the subject of this article, provide an independent tool to constrain local and regional carbonate sedimentary models and study their relationships with palaeoclimate and the carbon cycle (Totman Parrish and Curtis, 1982; Whalen, 1995; Kiessling et al., 2003).

In geology and specifically in seismic interpretation, deterministic and uncertain scientific approaches are applied to overcome data limitation (e.g., Michel et al., 2023). This data limitation is inherent to the study of natural systems and is extreme in the case of frontier exploration using seismic data alone. Deterministic models allow incorporating concepts into geologic interpretations and are based on the understanding of geologic processes and controlling factors (e.g., Borgomano et al., 2020). At a global scale, there is a strong relationship between geography and the distribution of carbonate platforms that has been illustrated in both the present day (Lees and Buller, 1972; Laugié et al., 2019) and the deep time (Davies et al., 2019; Pohl et al., 2019; Michel et al., 2020). At a more local scale, several studies have revealed a relationship between stratigraphic architecture and the type of carbonate producers and the geodynamic context (e.g., James and Clarke, 1997; Bosence, 2005; Tassy et al., 2023). The relative importance of these factors evolves through geologic time and can depend on palaeoclimatic, paleoceanographic, and geodynamic conditions, coupled or not to carbonate producing biota evolution (Markello et al., 2008; Pomar and Hallock, 2008). This time dependency implies that carbonate systems and reservoirs must be studied for specific geologic time (Kiessling et al., 2003). This study proposes a geologic expert system to predict the distribution and architecture of carbonate platforms from local to global scales and to support seismo-stratigraphic interpretations. The investigated stratigraphic interval is the Upper Jurassic, which hosts a significant portion of petroleum reserves, especially in the Middle East (Eltom et al., 2014), and lacks global carbonate predictive models despite abundant sedimentological and paleontological studies (e.g., Leinfelder, 2001; Martin-Garin et al., 2012).

In principle, expert systems consist in the transfer of expertise, which is the extensive set of specific knowledge possessed by an expert, from a human to a computer tool (Jackson, 1986) without implying artificial intelligence (Buchanan and Smith, 1988). This knowledge is stored in a numerical database that can be accessed by users to obtain specific advice through a set of facts and rules predefined by the expert (Bohanec and Rajkovi, 1990). Expert systems, which are conceptual in nature, can operate on an incomplete set of data and provide multiple solutions to a problem or query, with varying degrees of uncertainty (Shu-Hsien Liao, 2005). The use of expert systems in the study and prediction of geologic reservoirs and more generally in geosciences has already proven to be highly relevant (Nikravesh, 2004, 2007). The first expert systems emerged in the mid-1960s, but it was not until 1994, with the advent of the Internet opening to the general public, that there was a significant evolution of expert systems (Lukasheh et al., 2001). The expert system applies a multidisciplinary approach that integrates geophysical data (e.g., seismic images) and field sedimentological data (e.g., stratigraphic sections), along with a conceptual deterministic analysis and numerical modeling.

The implementation of the expert system applied to the Upper Jurassic carbonate platforms is structured in several steps (Figure 2). Step 1 is data collection (database). Quantitative data such as lateral dimensions, thickness, duration, and subsidence rate and qualitative information such as platform physiography, carbonate factory, and geodynamic context are extracted from the literature and integrated into a database.

Step 2 is data expertise (bibliographic knowledge base) in which a set of facts and rules are listed based on analyses of the database. The analyses aim to identify global trends in Upper Jurassic marine carbonate systems. These trends can result in geologic classes based on the most discriminative criteria for the global distribution of carbonate architectures, such as the geodynamic context or the carbonate factory, for example. Thus, the geologic class allows the estimation of parameters to a poorly documented case study.

Step 3 is process-based modeling (numerical knowledge base). Interpreted geologic processes, associated parameters, and concepts are tested by process-based stratigraphic modeling for each case study. Thus, the observed seismic images are reproduced to guide seismic interpretation.

Step 4 is carbonate prediction (inference engine). This functions like a search engine examining the knowledge base for information that matches the user's query. For example, a query related to a case study with very limited data, with a paleo-geographical location, will result in the prediction of stratigraphic and sedimentary parameters coherent with available knowledge such as the local geodynamic context. The case study will be assigned ranges of values for missing quantitative data and coherent qualitative parameters, such as seismic architecture images, thus providing knowledge and data to support the prediction.

For instance, an example extracted from the database and presented as a table (Table 1) will be implemented according to the different steps of the workflow (Figure 2) for generating three distinct database tables:

  • initial database table, including only parameters and information available in the literature;

  • expert and analogue database, with additional information and parameters resulting from the data analysis; and

  • prediction table, completed with information and parameters obtained from process-based stratigraphic modeling.

Data collection. Seventy-seven Upper Jurassic carbonate platform cases were documented through an extensive bibliographic search involving 205 documents, from both outcrops and subsurface. This literature survey focused on examples with specific description of shallow marine carbonate platform deposits (less than 100 m paleo-depth), combining information on dimensions, stratigraphy, age (according to the geologic time scale [GTS] 2022 chart), sedimentology, paleoecology, and carbonate factory.

The paleogeographic coordinates were calculated with the GPlates software (GPlates, n.d.,). The case studies were plotted on the paleogeographic maps of the Upper Jurassic (Scotese, 2014) allowing the analysis of paleogeography, paleoclimate, and geodynamic contexts (Figure 1).

For each case study, various quantitative and qualitative parameters are provided including the data source (outcrop or subsurface), precise location (coordinates and GPS paleocoordinates [World Geodetic System 84]), age (older and younger limits and duration), maximum dimensions (thickness, length, and width), preservation rate (ratio of maximum thickness to duration), geodynamic context (classification derived from Bosence, 2005; Tassy et al., 2023), connection with the continental domain (attached or isolated), morphology (platform or ramp), carbonate factory (Schlager, 2000, 2003, 2005; Michel et al., 2018), characteristic depositional facies including those of the platform margin, and corresponding bibliographic references. Due to data availability, some of the parameters could not be provided for all case studies; thus, out of 77 case studies, 49 have data on maximum thickness, 49 on geodynamic context, 49 on carbonate factory, and 36 on sedimentary profile. An image of the stratigraphic architecture has been created for the best-constrained cases (n = 10) (Figure 3).

Geologic data reported in papers are often incomplete and biased by deterministic interpretations and poor preservation (Michel et al., 2020). Furthermore, when working with data sets with drastically different levels of data support, it is common practice to generate identical conceptual models, but it is difficult to weight for purpose interpretations and conceptual models. This practice does not often distinguish model-driven from data-driven interpretations and predictions (Figure 4). This data collection phase, completing parameters for each case study, aims to standardize and normalize the data set for further analyses and studies.

Data expertise and analysis. The analysis of the database allows the description of the spatial distribution and associated parameters of the case studies. Forty-nine case studies were analyzed, out of 77; 28 case studies did not display enough data or showed too many uncertainties to be included in the statistical analysis. The objective of this analysis is to identify predictive geologic trends by categorizing the case studies into classes based on specific quantitative and qualitative parameter values. The case studies are analyzed using deterministic and conceptual classifications.

A wide variety of Upper Jurassic carbonate platforms is observed at global scale. The only spatial pattern that can be established at the global scale is the preferential palaeolatitudinal distribution of Upper Jurassic carbonate platforms between latitude 30° N and 30° S. The depositional durations vary from 2.2 to 21.1 Ma with a median value of 11.4 Ma, according to the GTS 2022 chronostratigraphic chart. A great diversity of maximum thickness is recorded ranging from 61 to 1900 m with a median value of 340 m. Not having data on thickness compaction, it is not considered in this study. The apparent accumulation rates (sensu Borgomano et al., 2020) or preservation rates (m/Ma) (Miall, 2016) are obtained by dividing the maximum thickness (m) by the depositional duration (Ma). These rates range from approximately 7 m/Ma to more than 142 m/Ma with a median value of 40 m/Ma (Figure 5).

Geodynamic context. The case studies were categorized according to their geodynamic context in accordance with the classifications of Bosence (2005) and Tassy et al. (2023). Three geodynamic contexts are considered — the extensive system (passive margin, syn-rift, and post-rift), intra/peri-cratonic system, and compressive system (foreland and back-arc basins). Out of the 49 case studies, 28 belong to the extensive/passive system, 14 to the intra/peri-cratonic system, and seven to the compressive system. The extensive/passive context exhibits the greatest thicknesses with a median of 520 m (255–1040 m), while the compressive and intra/peri-cratonic contexts show median values of 300 m (190–465 m) and 180 m (145–275 m), respectively. Only one case study of both intra/peri-cratonic and compressive context exceeds 13 Ma of duration, while the extensive context includes 12 out of 28 case studies lasting more than 13 Ma. These observations result in lower preservation rates for case studies associated with intra/peri-cratonic and compressive contexts compared to the extensive/passive context.

Carbonate factory association. This classification allows differentiating carbonate platforms based on characteristic carbonate producer types (James and Clarke, 1997; Schlager, 2000, 2005; Michel et al., 2018). The facies associations are relatively similar for all case studies corresponding to tropical factories–photozoan associations, except for the Vaca Muerta Formation in Argentina, where carbonate producers are associated with the cool-water factory–heterozoan association (James and Clarke, 1997; Schlager, 2000, 2005; Michel et al., 2018). The term “tropical factory” refers to carbonate sediments dominated by phototrophic grains (i.e., utilizing light as energy source) that are mainly produced by algae and symbiotic animals such as certain corals, foraminifera, and mollusks (Michel et al., 2018). These biota allow producing flat-topped platform closely related to the sea level.

Facies model. The typical sedimentary profile of the Upper Jurassic is characterized by (1) mixed siliciclastic-carbonate facies or evaporites in a proximal or coastal context; (2) facies rich in peloids, oncoids, and bioclasts in an internal platform context; (3) a more or less pronounced topographic external margin characterized either by bioconstructed reefs rich in corals and sponges, potentially accompanied by microbialites, or by an oolitic or bioclastic shoal, locally punctuated by coral and sponge mounds; (4) facies rich in intraclasts and bioclasts at the top of the slope, originating from the internal and external platform; and (5) marls and clays at the base of the slope and the basin (Wilson, 1975; Leinfelder, 1992; Kiessling et al., 2003). The trend toward uniformity of facies models is observed on a global scale, with the exception of the Vaca Muerta Formation in Argentina, which is located below latitude 30° S.

Platform morphology. Out of the 49 Upper Jurassic case studies, 36 sedimentary profiles were classified into ramps (n = 16) and platforms (n = 20), the latter being further subdivided into platforms with reef margins (n = 12) and nonreefal margins (n = 8) (Pomar, 2001). The statistical analyses reveal differences in maximum thicknesses and depositional durations between the different platform morphologies. Thus, the median values of maximum thicknesses and depositional durations are 1000 m and 15 Ma for platforms with reef margins, 350 m and 12 Ma for platforms with nonreefal margins, and 250 m and 9 Ma for ramps, respectively (Figure 6).

Stratigraphic architecture. Among the 10 carbonate platforms with well-established stratigraphic architectures, six are aggradational and four are progradational. Progradation distance varies from 300 m (Abenaki platform, Canada) to 165 km (Arab Formation, Saudi Arabia). Platforms developed in extensive geodynamic context either demonstrate an aggradational architecture or a short progradation of a few hundred meters. On the other hand, case studies with a relatively significant progradational architecture (several thousand meters) developed in either intra/peri-cratonic (Jura platform, France-Switzerland, and Arab Formation, Saudi Arabia) or compressive contexts (Vaca Muerta Formation, Argentina) (Figure 7).

All the cases are classified and grouped according to the information and parameter values obtained from the database analysis. Different ranges of parameter values are attributed to each class. The value ranges are defined by the first and third quartiles of the case study population of the considered geologic class. The value ranges of different classes — e.g., geodynamic, and morphological–facies association classes — can be combined by intersection when relevant. This classification system allows the prediction of parameter values (e.g., thickness) for poorly constrained case studies. Missing values will be replaced by the corresponding parameter values of the analogue class of the case study (Figure 8).

The expert and analogue database is completed according to the geologic classes and the ranges of parameter values filling the data gaps of poorly constrained case studies. The expert system can therefore compensate missing parameter values and complete the database in a traceable manner based on the geologic classes and analysis of existing data. For instance, in the case of the Barreiro Formation (Portugal) of which the maximum thickness is unknown but the geologic classes are interpreted as extensive system and reefal margin platform, the expert system will assign the corresponding thickness range of the reefal margin platform geologic class within the extensive system geodynamic context (255–1040 m), hence 700–1500 m (Figure 8).

Process-based modeling. Process-based stratigraphic modeling is introduced in the presented workflow to test geologic hypotheses at regional and local scales and predict the global distribution of stratigraphic architectures of carbonate systems (Burgess et al., 2006; Lanteaume et al., 2018). The process-based stratigraphic modeling software used in this study is DionisosFlow, developed by Beicip-Franlab (Granjeon, 1996; Granjeon and Joseph, 1999). This type of numerical modeling allows different scenarios to be tested by varying input parameters representing physical or ecological processes controlling the carbonate systems, such as eustatic variations, tectonic subsidence, or carbonate production (Borgomano et al., 2020). Given that geologic data are often biased due to their poor accessibility and preservation over time and space, process-based stratigraphic modeling helps to overcome these data limitations by providing valuable insights into missing information (Miall and Miall, 2001).

Process-based stratigraphic modeling is employed to test the physical consistency of accommodation scenarios using classical carbonate production models to reproduce the gross — i.e. thicknesses, overall aggrading or prograding, and overall facies distribution — stratigraphic architecture of the best-constrained case studies of the database (Husinec and Fred Read, 2007). The different simulations are then compared to validate the geologic classes and associated trends or parameters (Burgess, 2001; Burgess and Wright, 2003). The accommodation is computed with the Haq sea level curve (Haq et al., 1987), with low-frequency variations and subsidence estimated from the carbonate platform thickness. Carbonate production schemes (along sedimentary profile) defined as platform margin (corals, ooids, etc.) producing more than the lagoon and the slope (Wilson, 1975; Schlager, 2005), and sediment transport is more significant at platform margin. Carbonate production rates are set to match accommodation rate and realize the overall stratigraphic architecture (e.g., aggrading or prograding architecture). This forward modeling workflow can help to detect and estimate the parameter variations from the different geologic classes. For example, carbonate platforms from extensive/passive settings show subsidence rates three times higher (65–95 m/Ma) than the rates of systems in intra/peri-cratonic contexts (20–30 m/Ma). Such simple but fundamental patterns are validated as a predictive rule, which can be used for predicting ranges of values for poorly constrained case studies according to their geologic classes (Figure 9).

The Barreiro Formation, whose thickness and stratigraphic architecture are unknown, is a good example of application of this workflow. This case study was selected due to the lack of available data, especially the absence of seismic data. Additionally, the Barreiro Formation is located in a geographic region where we have some well-documented analogous cases, such as the Abenaki platforms (Canada) and Tarfaya (Morocco).

The palaeogeographic position of this carbonate platform along the northern Atlantic corresponds to an extensive geodynamic context (Wenke, 2014). Given the abundance of coral reefs at the platform margin and the morphology of a rimmed platform, the Barreiro Formation is classified as a reefal margin platform (Ellis et al., 1986). The estimated duration is 11.4 million years (Oxfordian–Kimmeridgian). According to the geologic class reefal margin platform associated to an extensive/passive geodynamic context, the expert system predicts a thickness range of 700–1500 m and an aggrading stratigraphic architecture. These predicted parameters are used to constrain the forward modeling. A smooth initial topography with a low angle slope forming a distally steepened platform is based on analogous platforms (i.e., Abenaki platform, Canada, and Tarfaya platform, Morocco) from the same palaeogeographic region and geodynamic setting (Central Atlantic Ocean margins). The sedimentation rates of the different facies associations used to model the Barreiro Formation are derived from simulations of case studies belonging to the same geologic group, i.e., reefal margin platform (see 1, 2, 3, and 4 from Figure 9). The subsidence rates are directly constrained by the predicted thickness and aggrading stratigraphic architecture and are calibrated through modeling iterations together with sedimentation rates. The two stratigraphic forward 3D models represent endmembers of the Barreiro platform constrained by the predicted ranges of thickness and production rates (Figure 10).

The use in stratigraphic forward modeling of production rate ranges attributed to the geologic class reefal margin platform in extensive/passive geodynamic contexts, result in thickness ranges and stratigraphic architectures consistent with available data and knowledge. The two endmember models are created respectively with the lowest and highest production rates in combination with variable subsidence rates to obtain aggradational architectures. The thickness range of the Barreiro platform at the end of this iteration is 720–1140 m (Figure 10).

Global carbonate platform prediction. The previous analyses allow the development of a predictive global palaeogeographic map of Upper Jurassic carbonate platforms and geodynamic regions (Figure 11a). A level of data constraint is assigned to each carbonate platform case. The contouring of geodynamic regions distinguishing extensive/passive, intra/peri-cratonic, and compressive geodynamic regimes, is based on the case study documentation and published palaeogeographic maps (e.g., Scotese, 2014). For each geodynamic region, a general stratigraphic architecture and a range of thickness values can be established. For poorly constrained case studies such as frontier exploration seismic lines, the expert system is hence able to purpose predictions of stratigraphic architectures and ranges of thickness depending on the geologic classes including the geodynamic region (Figure 11b).

Using a process-based, global-scale method and applying a multidisciplinary approach that integrates geophysical and geologic data, the geologic expert system can provide standardized predictions of carbonate stratigraphic architecture (Figure 11). The expert system methodology is supported by case-by-case bibliographic studies, which represent meticulous and objective observations from the field. Through analogue comparison and classification of global trends, the expert system operates a replication of data to form hypotheses and theories (cf. Frodeman, 1995). Such an approach helps to overcome the inherent complexity of geologic investigations, which are carried out from incomplete records and thus do not deal with the entire history and composition of a sedimentary system (e.g., Kidwell and Bosence, 1991; Wright et al., 2003; Cherns et al., 2011). The deterministic approach, which is based on geologic concepts and classifications, methodologically tackles the fact that there are no theory-free facts in geology (cf. Rudwick, 1996). As an illustration of this latter statement, thickness and facies data are meaningless without considering preservation and facies model concepts, respectively. Hence, the expert system relates thicknesses and facies data to accommodation (i.e., geodynamic) contexts (e.g., Tassy et al., 2023) and carbonate production models, which are, in turn, related with predictable paleoceanographic conditions (e.g., carbonate factory concept; Schlager, 2005; Michel et al., 2018). In addition, the expert system utilizes numerical stratigraphic forward modeling as an experimentation tool to overcome the difficulty of reproducing geologic objects through analogic experimentation, given the long-time span at which geologic processes operate and their great dimensions.

The expert system hence combines different approaches. First, the analogue database comes from case-by-case studies from the field. The case-by-case approach involves a careful and in-depth examination of available data and identifies similarities with other case studies to establish general rules. Even though the observation may seem impartial, the case-by-case approach can still lead to divergent interpretations and controversies. This problem originates from the analysis of rocks, which record incomplete geologic evidence of the paleoenvironmental processes governing the original natural system, thus leaving a large space for geologic speculations. Controversies also stem from human factors and biases, as geologists are often influenced by their past experiences, schools of thought, and other forms of influence that have marked their career path (Miall and Miall, 2001). This approach, where one starts from a specific case study to search for similarities and establish generalizations, follows an inductive reasoning and does not involve experimentation in a majority of cases. In such a context, the expert system integrates complementary approaches to better constrain and test concepts and interpretations, which is crucial to the geologic practice.

Then, the expert system focuses on the accumulation of many case studies that will undergo thorough analysis and comparison to consolidate general trends. Each case study presents varying data sets with some being well constrained, while others are much less so. To ensure consistent and reproducible comparison between the various case studies, standardization is imperative. This process involves providing predefined parameters for each case study such as maximum thickness and geodynamic context. The standardization step also minimizes human biases in framing the interpretation of case studies. From the analyses and comparisons, geologic hypotheses are formulated to highlight trends and establish predictive rules regarding the overall distribution and stratigraphic architecture patterns of Upper Jurassic carbonate platforms.

Finally, stratigraphic forward modeling is used as a numerical geologic experimentation (Burgess et al., 2006; Lanteaume et al., 2018; Borgomano et al., 2020). Numerical modeling allows simulation of complex processes such as, for example, stratigraphic accumulation under varying accommodation scenarios and climate changes (Williams et al., 2011). Modeling can then be used to test concepts and interpretations in a physically consistent multidimensional space (Figure 11) (Montaggioni et al., 2015; Lanteaume et al., 2018). Hence, the expert system methodology clarifies the scientific approach of hypothesis formulation and testing attempting to limit the effect of geologist preconceptions, which often influence observations and interpretations. The system expert methodology establishes a loop process coupling inductive (i.e.: case-by-case bibliographic study) and deductive reasoning (i.e., process-based, global-scale geologic classifications), and includes numerical experimentation.

In practice, the expert system provides a consistent and standardized knowledge base that allows contextualizing any Upper Jurassic carbonate platform. The expert system further provides information about the global controlling factors of Upper Jurassic carbonate stratigraphic architecture including ranges of predictive parameter values. Additionally, the expert system invites the geologist to test hypotheses using numerical modeling, which in turn will provide a catalog of carbonate stratigraphic architectures (Figure 9).

Carbonate producers. During the Upper Jurassic, the depositional facies associated to carbonate platforms show a relatively homogeneous global distribution dominated by carbonate producers of the T-factory (Figure 6) (James and Clarke, 1997; Michel et al., 2018; Schlager, 2000, 2005). Except for the Vaca Muerta, which corresponds to a C-factory located in the highest paleolatitudes (Iglesia Llanos et al., 2017; Rodriguez Blanco et al., 2020), all the other carbonate platforms are located at a paleolatitude lower than 30° S. Moreover, most Upper Jurassic carbonate platforms (74 out of 77 case studies) developed between the two 17°C isotherms (Scotese, 2014). There could be a palaeoclimatic trend in the distribution of carbonate producer types beyond the 30° S palaeolatitude and 17°C paleotemperature (Figure 11). This relative sedimentary homogeneity aligns with the global palaeoclimatic context during the Upper Jurassic, which is characterized by relatively warm and humid conditions (Hallam, 1984; Hallam et al., 1993; Rees et al., 2004) and extends across a wide tropical and subtropical region reaching up to around 40° N (Moore et al., 1992; Jenkyns et al., 2012; Korte et al., 2015). In this context, it remains unknown whether the Vaca Muerta is part of a palaeoclimatic trend, or it represents a peculiar case study in which terrigenous and local tectonic influence impacts the carbonate factory.

In the context of more detailed sedimentary profiles, Upper Jurassic facies distribution can be consistently interpreted as facies belts that match the different depositional environments (Figure 6) (Wilson, 1975; Leinfelder, 1992). Proximal depositional environments include coastal areas rich in siliciclastics and inner parts of platforms, which are mainly composed of peloids, oncoids, and bioclasts. The more distal outer platform is composed of granular oolitic or bioclastic shoals and may include coral bioconstructions often associated with sponges. It is worth noting that coral-sponge bioconstructions can be found on the sedimentary profile either in patch form (e.g., Arab Formation) or as continuous vertical deposits that can act as barriers between the outer and inner parts of the platform (e.g., Abenaki platform). The platform slope exhibits very coarse facies composed of intraclasts and bioclasts in the upper part due to the transport of grains from both the inner and outer platforms, and finer marl and clay facies in the more distal part. This arrangement of depositional environments is observed in all palaeogeographic zones worldwide (Leinfelder, 1992; Kiessling, 2009; Martin-Garin et al., 2012). Microbial deposits can be found throughout the sedimentary profile and hence do not represent a discriminating paleoenvironmental proxy (Dupraz and Strasser, 1999; Olivier et al., 2012; Ricci et al., 2018).

Sea-level variations. Geodynamic changes such as episodes of oceanic ridge uplift and subsidence, along with palaeoclimatic variations, resulted in eustatic variations (Hallam, 1978) materialized by a sea level rise of 36 m between the Oxfordian and the end of the Tithonian (Haq et al., 1987). The stratigraphic forward modeling (Ch. 2.3) demonstrates that the Haq sea level curve has no particular impact on the stratigraphic architectures of the Upper Jurassic carbonate platforms. This important result can be explained by the low amplitude and frequency of the sea-level variations (36 m) compared to the high ranges of subsidence exceeding several hundred meters (Figure 9). This is consistent with the interpretation of reduced sea-level variations during greenhouse periods (e.g., Read, 1998), which are characterized by sea-level fluctuations of a few meters during both periods of 10–20 ka and 100–400 ka sequences (Koerschner and Read, 1989; Goldhammer et al., 1993). On the scale of Upper Jurassic carbonate platforms, such medium and small-scale sequences were identified by several authors in various formations (e.g., the Adriatic and Jura platform). These sequences were associated with sea-level fluctuations induced by orbital eccentricity cycles on periods of 100 and 400 ka (Dupraz and Strasser, 1999; Strasser and Samankassou, 2003; Husinec and Fred Read, 2007; Védrine and Strasser, 2009; Cariou et al., 2014). In our study, however, we were unable to address these first- and fourth-order sea-level fluctuations because too few case studies provided information on the record of such cyclic patterns and because chronostratigraphy is poorly constrained. During the most pronounced transgressive and regressive patterns of the Haq curve (Haq et al., 1987) during the Upper Jurassic, occurring during the intervals 152–158 Ma and 150–146 Ma, respectively, no significant change is observed in the simulated models (Figure 9). This lack of signal is likely masked by subsidence, for which rates are two to eight times greater than eustatic variation rates. Higher frequency than Haq sea-level curves could be tested in forward modeling. The lack of such a eustatic curve and of age control, however, prevent any well-constrained analysis. In any case, sea-level variations alone, which represent a unique, global signal, cannot explain neither regional nor local differences in stratigraphic architectures.

Geodynamic contexts. There is a great variability of maximum thickness records of Upper Jurassic carbonate platforms from 60 to 1900 m. The palaeogeographic distribution of maximum thickness as well as overall stratigraphic architectures can be related to the geodynamic contexts, which often represents an understudied controlling factor of carbonate platform development (Bosence, 2005; Tassy et al., 2023). Extensive/passive geodynamic contexts exhibit the greatest thicknesses and preservation rate of carbonate platforms (255–1040 m; 25–80 m/Ma), whereas compressive (190–460 m; 25–50 m/Ma) and intra/peri-cratonic contexts (145–275 m; 20–40 m/Ma) show relatively lower thicknesses (Figure 8). The thickness trend is directly associated with a subsidence trend: extensive systems showing greater subsidence rates and height than intra/peri-cratonic systems (see Xie and Heller, 2009). If compressive systems can show large subsidence patterns, such systems appear not to be the most favorable context for carbonate platform development, which might be related to relatively unstable paleoenvironmental conditions and large terrigenous input (Bosence, 2005; Tassy et al., 2023).

Two types of stratigraphic architectures were observed in our study: aggradational and progradational architectures (Figures 7 and 12). All case studies exhibiting a progradational architecture developed in either an intra/peri-cratonic (n = 2) or compressive (n = 1) geodynamic context. Aggradational architectures were observed in case studies belonging to the extensive/passive geodynamic context (n = 5), except for the Abenaki platform (Canada), which exhibited a weak progradation considered nonsignificant at the platform scale (approximately 300 m), and the Kugitang Formation (Uzbekistan) belonging to the intra/peri-cratonic geodynamic context. However, it is worth noting that information regarding the platform margin in the case of the Kugitang Formation is not documented due to the lack of outcrop preservation. These classes of progradational versus aggradational architectures can be related to the observation that the lowest accommodation rates were found in intra/peri-cratonic and compressive contexts (cf. above preservation rates, which reflect accommodation trends), while the highest rates were found in the extensive/passive geodynamic group. Thus, for the same carbonate production, systems with relatively low accommodation such as intra/peri-cratonic and compressive contexts tend to prograde, while aggradation occurs in an extensive/passive context where accommodation rates are higher.

The study conducted by Tassy et al. (2023) dealing with the geodynamic trends of carbonate platforms throughout the Phanerozoic reveals that case studies from the Lower-Middle Jurassic primarily developed in rifting-drifting contexts (five out of eight case studies) and do show progradation (Figure 12). In contrast, case studies from the Cretaceous mainly developed in passive margin contexts (13 out of 19 case studies) exhibiting an average progradation of 23,000 m. The preservation rates, which are akin to accommodation rates, are higher for the Lower-Middle Jurassic (median value of about 50 m/Ma) compared to the ones of the Cretaceous (median value rate of about 30 m/Ma). Geodynamic contexts of rifting-drifting are interpreted to be associated with higher subsidence rates favoring aggradational architectures, while the passive margins of the Cretaceous experience less subsidence resulting in the development of progradational carbonate platforms. No trend is observed regarding the maximum thicknesses of the Lower-Middle Jurassic and Cretaceous case studies.

The Jurassic case studies, covering all stages, show higher preservation rates than the Cretaceous case studies. The thickest sections are observed in the Upper Jurassic (with a median of 1250 m), while the thicknesses are lower and appear similar for both Lower-Middle Jurassic and Cretaceous periods (with a median of about 300 m) (Figure 12). These differences in preservation rates between the Jurassic and the Cretaceous could be explained by the geodynamic contexts. During the Jurassic, the breakup of Pangaea led to an opening phase, promoting the development of rifting/drifting systems on a global scale and characterized by more significant subsidence compared to the largely dominant passive margin contexts during the Cretaceous (Steuber, 2002; Pohl et al., 2020). Although passive margins are found in the Upper Jurassic, they are younger and likely still influenced by a rifting/drifting geodynamic setting (Wenke, 2014). This difference between younger passive margins in the Upper Jurassic and relatively older Cretaceous passive margins could explain the differences in carbonate platform thicknesses.

The variation in thickness between the Upper Jurassic and the Lower-Middle Jurassic, despite a similar geodynamic context and a prevailing aggradational architecture, could be due to differences in production rates. According to the literature, the production rates of the Middle Jurassic are considered to be low (decreased by approximately 20% to 40% compared to the Cretaceous for example) due to significant global cooling (Dromart, 2003; Donnadieu et al., 2011). The increased temperatures in the Upper Jurassic, associated with increased carbonate production rates, might result in increased thicknesses for carbonate platforms.

Thus, in the Lower-Middle Jurassic, high accommodation rates and lower production rates might contribute to aggradational and relatively thin carbonate platforms. In the Upper Jurassic, high accommodation and carbonate production rates favor the development of aggradational and relatively thick architectures. Moderate accommodation rates and high production rates in the Cretaceous facilitate the development of progradational and relatively thin carbonate platforms. Note that in the present study, the 2D observations reveal simple predictive trends. For a more detailed understanding of the different stratigraphic architectures and rates, 3D volume estimations could be valuable.

Spatial variations. During the Upper Jurassic, carbonate platforms occupied broad bands of the Tethyan margins between 40° N and 40° S of palaeolatitude and within 17°C isotherms due to a greenhouse palaeoclimate (Moore et al., 1992; Hallam et al., 1993; Weissert and Mohr, 1996). The favorable palaeoclimate for carbonate platform development was coupled with (1) a relatively high sea level — approximately 100 m above the present sea level — and (2) an extensive global geodynamic context initiated by the fragmentation of the Pangaea supercontinent (Pillevuit et al., 1997; Stampfli and Borel, 2002; Scotese, 2014; Tassy et al., 2023). These palaeoenvironmental conditions offered large available space for Upper Jurassic carbonate accumulations.

Case studies showing the highest thicknesses and associated with the highest accommodation rates (greater than 60 m/Ma) predominantly occur in extensive and passive contexts and exhibit aggradational architectures — e.g., Abenaki, Tarfaya, Apulia, and Adriatic platforms — while case studies with relatively low accommodation rates (less than 40 m/Ma), are mainly found in intra/peri-cratonic and compressive contexts and display progradational architectures — e.g., Arabian Formation and Jura platform. Since the same types of carbonate producers of the T-factory are found in all sedimentary profiles (Leinfelder, 1992, 2001; Kiessling, 2009), carbonate production rates are considered to be relatively comparable for each case study. Furthermore, greenhouse periods promote the constant filling of available space due to low eustatic variations (Read, 1998; Husinec and Fred Read, 2007). Hence, tectonic subsidence, which is predicted by regional geodynamic contexts, is interpreted as the limiting factor of maximum carbonate thicknesses and as the steering controlling factor of overall aggradational and progradational patterns.

Even though carbonate producers are relatively similar worldwide, various types of ramp and platform profiles co-exist during the Upper Jurassic period. These distinct morphologies were investigated in relation to geodynamic contexts to observe if there were any potential trends concerning thickness values (Figure 8). It appears that ramp-type profiles do not exhibit significant differentiation based on the geodynamic context (from 150 to 500 m). Carbonate platforms of the reefal margin platform type show greater platform growth potential than nonreefal margin platforms. The greatest thicknesses are observed in reefal margin platform profiles developing in an extensive/passive context (700–1500 m) and to a lesser extent in a compressive context (350–700 m). It is noteworthy that no case studies of reefal margin platform types developing in an intra/peri-cratonic context were recorded. Nonreefal margin platform platforms are found in all geodynamic contexts; however, the highest thicknesses are observed in extensive/passive contexts (300–1100 m), while the lowest thicknesses are observed in intra/peri-cratonic and compressive contexts (180–250 m).

The proxy of maximum thickness of carbonate architectures shows some limitations. This parameter, which does not consider preservation biases, only relates to global-scale, first-order geologic processes such as large-scale accommodation trends. Thickness is also only partially related to carbonate production because it refers to the aggradational component of carbonate platforms. Progradational patterns are excluded from this proxy. Estimations of carbonate volumes, especially progradations, could be estimated by 2D cross-referencing maximum thickness with lateral distances (Montaggioni, 2005; Michel et al., 2020). In the present study, the weak patterns of Upper Jurassic carbonate progradations limited the global signal of carbonate volume. Despite these limitations, the maximum thickness parameter displays coherent global trends in terms of a geologic concept. Therefore, we use this parameter to formulate predictions regarding the distribution of Upper Jurassic carbonate architectures.

This study established and implemented a methodology based on an expert system to predict the distribution of Upper Jurassic carbonate platform architectures on a global scale. The expert system supports seismo-stratigraphic interpretations, especially in frontier exploration contexts. Through the collection and standardization of data, the analysis of data and parameter values, and the creation of associated geologic classes, the method effectively addresses the data gaps of case studies. The data analyses, coupled with the use of process-based modeling, revealed spatiotemporal trends in Upper Jurassic carbonate platforms, enabling predictions of the global distribution of architectures. The expert system can support interpretation of seismic data given a paleogeographic location and an estimated age interval of a case study. A catalog of typical carbonate platform architectures is proposed to improve seismic interpretation.

This study also contributed to enhancing our understanding of the Upper Jurassic by revealing several significant observations:

  1. the uniformity of carbonate producers worldwide;

  2. the limited impact of eustatic changes on architectures due to their relatively moderate amplitude, considering the substantial thicknesses of carbonate platforms observed;

  3. the substantial influence of geodynamics on architectures, with the thickest carbonate platforms occurring in extensive/passive tectonic settings (median = 500 m), while thinner platforms develop in intra/peri-cratonic contexts (median = 180 m);

  4. the importance of morphological factors to some extent, particularly for platforms characterized by continuous vertical coral buildups, which stand out distinctly from other morphological types — i.e., the median values of the geologic classes of reefal margin platforms, nonreefal margin platforms, and ramp equal 1000, 350, and 280 m, respectively.

Thus, the use of an expert system in the study of carbonate rocks proves to be an extremely effective tool for advancing our knowledge of global carbonate platforms, allowing a better prediction of the distribution of carbonate stratigraphic architectures worldwide.

We want to acknowledge AKKODIS for permission to publish this paper and Beicip-Franlab for providing the DionisosFlow software that was used for the stratigraphic forward modeling for this study.

Data associated with this research are available and can be obtained by contacting the corresponding author.

All article content, except where otherwise noted (including republished material), is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND) license. See https://creativecommons.org/licenses/by-nc-nd/4.0/. Distribution or reproduction of this work in whole or in part requires full attribution of the original publication, including its digital object identifier (DOI). Commercial reuse and distribution of derivatives are not permitted.