Defining electrofacies models in carbonate reservoirs is challenging due to well-log resolution limitations and diagenesis, among other factors. Although electrofacies have been successfully applied for large-scale reservoir characterization, their effectiveness at finer scales is limited by carbonate heterogeneity, varying log resolutions, and the influence of diagenesis on log responses. Shoulder bed effects resulting from well-log resolution can introduce artifacts that misrepresent geologic features. This increases uncertainty in predicting geologic information through electrofacies, particularly in heterogeneous, diagenetically altered carbonate reservoirs. This study addresses the challenges of electrofacies classification in the Aptian Barra Velha Formation, Santos Basin, Brazil, highlighting how shoulder bed effects create misleading interpretations. We developed a methodology that improves electrofacies modeling by (1) eliminating potential artifacts from shoulder bed effects and (2) constructing models using logs with similar resolutions and geologic significance to reduce misinterpretations through diagenetic alteration. Although we illustrate this approach using multiresolution graph-based clustering, the core innovation lies in providing greater interpretative control over electrofacies modeling and validation. The results demonstrate that this workflow increases confidence in predicted electrofacies representing resolvable rock intervals while increasing the understanding of their geologic meaning.
