Summary This paper introduces Geo-RAG, a customized retrieval augmented generation (RAG) framework that utilizes large language models (LLMs) for the digital transformation of unstructured geological documents, offering a more efficient and accurate method for extracting insights. The integration of advanced AI technologies in the Geo-RAG framework marks a significant advancement in geological document processing and information retrieval, demonstrating the potential for substantial efficiency gains and improvements in accuracy in the domain of earth science.
Summary Traditional petrographic analysis of rock thin sections is time consuming, it is potentially subjective, may suffer from operator bias, and the results are not necessarily representative of the bulk rock sample. The challenge to automation of petrographic description is the complexity that is inherent to geology; sedimentary rocks often contain many types of grains, cements, and pores. Here, we present a machine learning solution to automatically characterize minerals and grains in rock thin sections on a whole slide basis. The machine learning-based image analysis approaches outlined in this paper are designed for extraction of first-order rock properties at scale, unlocking latent data from 100s–1000s thin sections per project, including new or legacy slides. This is made possible in part by leveraging large pretrained vision models that can perform segmentation tasks on unseen images without additional training. The presented method also capably predicts most major minerals on a whole slide basis, although some limitations exist. Thin section screening through machine learning enables (1) more informed, data-driven, faster decisions, and (2) allocation of time and resources for specific, targeted microscale sample challenges which are best addressed using advanced imaging techniques such as QEMSCAN, and/or through inspection by a subject matter expert.
Summary Elastic full-waveform inversion (FWI) is now emerging as an industrial tool for the compressional velocity (Vp) model build, driven mainly by diving and reflected P waves. On the other hand, the inversion of S waves in elastic FWI to update shear velocity (Vs) models is more challenging due to the limitations of data acquisition and reduced sensitivity of surface seismic to Vs. In this paper, we propose a practical methodology for low-wavenumber Vs updates by elastic FWI, driven mainly by converted waves in multi-component ocean-bottom seismic data. The first key step of our methodology builds a high-quality Vp model from elastic Vp FWI using the hydrophone and vertical geophone data. The second key step is the use of horizontal geophones to reconstruct the low wavenumbers of the Vs model from the kinematics of the converted waves. We show a field application of our approach highlighting improved PS reverse time migration (RTM) imaging, and better consistency with the PP RTM, coming from the elastic Vs FWI. In addition, a Vs update from a Born-based PS-reflection FWI method also shows a good PS RTM uplift, although the elastic FWI model generated the better PS RTM image.
Summary Since the discovery of pre-salt oil fields offshore Brazil two decades ago, the Santos Basin has transitioned from an exploratory to a development phase. This requires a more comprehensive understanding of the reservoir physical properties. This study aims to demonstrate, on one of the biggest Brazilian oil fields, how five years of technological evolution for OBN processing can improve our understanding of the pre-salt reservoir. The Tupi producing field presents various imaging challenges due to its thick and stratified salt and complex pre-salt layer. The latest technologies, such as elastic Time-Lag FWI, RTM angle gathers using spherical binning and internal multiple attenuation, provide a more accurate velocity model, as well as better fault definition and AVA response. The impact on reservoir-based inversion of the Vp/Vs ratio shows decreased uncertainty at the target level.
Abstract This introduction briefly outlines the potential impact of igneous processes on sedimentary basins and their energy resources. The associated volume includes contributions across a range of scales; from margin-wide rifting to impacts on reservoir quality. The impact of igneous systems on elements such as tectonics, heat flow, hydrocarbon (and other) fluid charge, structuration and reservoirs are reviewed with a direct impact on hydrocarbon exploration and production, geothermal and carbon capture storage projects. A strong understanding of these commonly overlooked, igneous processes is likely to enable successful exploration for additional energy resources such as frontier hydrocarbons, geothermal heat, hydrogen as well as helium.
