Paper submitted to IMAGE 2026, by Patricia E. Rodrigues, Reinaldo J. Michelena(SeisPetro Geoconsulting), Keyla Gonzalez, Alejandro A. Valenciano (TGS).
Abstract
This work introduces a geology-anchored, two-stage workflow for large-scale log normalization designed to remove non-geologic variability while preserving stratigraphic and facies variations. The method integrates a systematically selected reference dataset with fully automated three-dimensional (3D) percentile-based normalization, enabling consistent basin-wide results at unprecedented scale across tens of thousands of wells. In the Midland Basin, where logs are affected by multiple casing intervals, different service companies, tool vintages, and sharp lateral facies transitions, traditional normalization approaches, focused on narrow target zones or neighboring wells, often fail to maintain geological and statistical consistency. Our workflow addresses these challenges through two stages: (1) reference normalization of high-quality wells using P10-P90 percentiles, and (2) generation of calibrated 3D statistical volumes that act as normalization engines for all remaining wells. Applied to more than 6,600 wells, the process achieved 98% successful normalization and was subsequently extended to over 20,000 additional wells using the same 3D volumes and rules, requiring only hours of additional computation once the volumes were built. The results show improved stratigraphic continuity, removal of casing-related artifacts, and clearer regional trends. Beyond Gamma Ray logs, the approach provides a scalable framework for normalizing other log types and preparing reliable inputs for large-scale property and facies modeling, seismic-log foundation models, and machine-learning workflows.
Read the full article here.