In deepwater fields, drilling costs are extremely high, and the "one sand, one well" situation is common, in which a single well must control an overly large area of the gas field. Structural accuracy decreases in areas distant from the well locations, making gas reservoir prediction and sand body delineation challenging due to the limited resolution of seismic data. Toaddressthesechallenges,thisstudy applied high-precision full-waveform inversion (FWI) velocity modeling and broadband imaging technology in a deepwater exploration of the South China Sea. In the preprocessing stage, based on the geological challenges and features of the acquired seismic data, we selected appropriate signal-processing methods and optimized the algorithms and parameter sets, successfully developing a customized broadband processing workflow specifically tailored for deepwater applications. The entire broadband processing sequence effectively supported subsequent FWI modeling. During the imaging stage, FWI was successfully applied for the 1st time in the deepwater of the South China Sea. Together with Q pre-stack depth migration, this integrated approach effectively addressed challenges in structural depth prediction and significantly improved imaging resolution. This study provided real-time support for gas field development and optimized the well placement for deepwater development.
This study evaluates the suitability of using existing subsea capping stack (SCS) designs to control CO2 blowouts. The physics behind controlling a methane (i.e., CH4) subsea blowout using an SCS is well understood by the oil and gas industry. The same, however, is not true for CO2, whose thermophysical properties differ greatly from CH4. Hence, the risks associated with halting a CO2 blowout with an SCS are not clearly understood. Understanding such risks, though, is an indispensable step in preparing for the future possibility of a subsea CO2 blowout. Numerical simulations of blowout/SCS scenarios for various combinations of water depth (300 or 762 m), blowout rate (4.25, 8.93 or 10.49 MMm3/d), reservoir fluid (CO2 or CH4), and SCS configuration (three variants) are conducted, with the goal of generating a broad spectrum of results for understanding the differences between installing an SCS on CO2 versus CH4 blowouts. Each simulation covers the full progression of a blowout (from initial reservoir fluid influx to final closure of the SCS). Multiphase computational fluid dynamic (CFD) modeling of the entire sequence is performed using a version of the OLGA model, newly applied to CO2 well-control events. High-rate CO2 blowouts in shallow water create conditions conducive to significant hydrate and ice formation, whereas CH4 blowouts do not (under any condition). The configuration of an SCS has little impact on outcomes. This suggests that using an SCS to control a shallow-water, high-rate CO2 blowout may encounter difficulties, implying that relief wells may be a necessary back-up option.
Many papers refer in a revised way to the two-step scenario of the Messinian Crisis conceived by Clauzon et al. (1996). The present paper recalls the basis for the two-step scenario and discrepancies with the later modified version, completed by new data supported by extensive micropaleontological analyses. Our interpretation of the Sicilian Eraclea Minoa section as belonging to a peripheral basin is the centre of the debate. We show the great amplitude of fluvial erosion during the peak of the crisis, which for the Rhône River, exceeded 400 km upstream of the present shoreline. Based on dinoflagellate cysts, we also recall the reasons for supporting the occurrence of three successive Lago Mare episodes of two different origins. The first and third episodes constitute phases of high sea-level exchanges between the Mediterranean and the Paratethys respectively just before the onset of paroxysm and after it. The second episode is due to overflowing Paratethyan waters from the Aegean Basin just before the end of paroxysm. Similarly, the demonstration of the marine reflooding of the Mediterranean Basin prior to the Zanclean is repeated. We emphasize dissimilarity between basins, focussing in particular on those, isolated or perched ones, which were continuously filled by waters during the desiccation phase: western part of the Alboran Sea and southeastern part of the Levantine Basin (marine waters), Apennine Foredeep (fresh waters), and Aegean Basin (brackish waters). The Apennine Foredeep cannot be the reference for the entire Mediterranean with respect to its evolution during the crisis. During the crisis, water exchanges between the Aegean Basin and the Eastern Paratethys (Dacic Basin, Black Sea) were impossible through the Marmara region because of the development of two opposed fluvial networks. Such exchanges existed thanks to a gateway that was probably located within the Balkans. Investigations around the Levantine Basin point to areas submitted to fluvial erosion during the crisis paroxysm and nearby areas, which might have received marine waters from the Red Sea. Much information is still to be discovered and that more progress is still needed in order to fully decipher this outstanding event.
This paper introduces a multi-agentic solution that leverages Generative AI—specifically, Large Language Models (LLMs) coupled with domain-specific engines—to enhance the efficiency, consistency, and technical depth of reservoir simulation workflows. The solution targets three high-value areas: simulation model compliance, insight generation, and well placement optimization, with the goal of accelerating field development planning and institutionalizing engineering best practices. The system is built around multiple AI agents, each integrating LLM-based natural language interfaces with specialized domain engines tailored for reservoir engineering tasks. The Reservoir Model Assessment Agent automates audits of simulation models by validating inputs, well constraints, and history matches internal modeling standards. The Reservoir Model Insights and Assessment Agent enables engineers to analyze, extract, and visualize critical model behaviors—such as production trends, scenario comparisons, and pressure evolution—through conversational queries. The Well Placement Optimization Agent blends simulation outputs with geospatial and operational constraints to generate ranked infill opportunities, simulate their performance impact, and provide rationale for each recommendation. All agents operate within a secure enterprise platform and are accessed via a unified, web-based interface that abstracts technical complexity while preserving engineering rigor. Applied to a synthetic reservoir with over 100 wells sourced from various publicly available sources and multi-decade production history, the multi-agentic system demonstrated substantial performance improvements. Model compliance reviews were completed in less than 20% of the time required by traditional manual methods, with the AI agent identifying more than 85% of known deviations. Insight extraction tasks that typically took 2–3 days were reduced to under an hour, while enabling deeper analyses such as recovery factor diagnostics and scenario benchmarking. The well placement agent rapidly evaluated over 25 new infill candidates, identifying zones with up to 6.5% projected ultimate recovery uplift. The domain-specific engines embedded in each agent ensured high fidelity in technical outputs, while the natural language interface enabled broader accessibility for both junior and experienced engineers. The deployment also standardized workflows across teams, reduced reliance on tacit knowledge, and improved transparency in decision-making. This work presents one of the first multi-agentic Generative AI solutions in reservoir engineering, combining the flexibility of LLMs with structured domain engines to deliver intelligent, explainable support across key simulation workflows. The result is a scalable, engineer-centric system that bridges AI and subsurface science to accelerate FDP and improve technical quality.
