Visualizing Geoscience Data Interpretation Into Photorealistic Modern Analog Using Generative AI: A Preliminary Result From Carbonate Platform Environment

Abstract

Abstract Geoscience datasets are fundamental for subsurface investigation. Paradoxically, they are sometimes exclusive and require subject-specific expertise to interpret and visualize. One such example is seismic interpretation. Geophysicists typically reconstruct ancient depositional settings by interpreting a myriad of seismic attributes and drawing analogs to the sedimentary process of the modern depositional environment (Posamentier et al. 2007; Vahrenkamp et al. 2019; Ramdani et al. 2021). Most of these interpretations will likely be reflection amplitude, frequency, impedance, or other geophysical attributes interpreted and "visualized" in the present-day geomorphology context (Posamentier et al. 2007; Warrlich et al. 2019; Ramdani et al. 2022b). The interpreter will then rely on verbal or written descriptions to convey their interpretation. Often, these descriptions are only well understood by fellow interpreters. Attempting to convey the same interpretation to a non-expert requires some degree of visual aid. Thus, a method to picture geophysical signals as a "depositional environment" is needed to bridge this gap. This study aims to leverage the application of generative AI as a tool for seismic interpretation. We propose a Conditional Generative Adversarial Network (CGAN)-based methodology capable of converting seismic attribute maps into photorealistic images of the modern satellite imagery analog as a visual aid for seismic interpretation.