Adaptive UMAP based multiple-frequency attribute blending and its application on hydrocarbon reservoir characterization

Multiple-frequency attribute blending analysis has become a significant tool for characterizing hydrocarbon reservoirs. Researchers commonly use time-frequency transformations to obtain multi-frequency components of seismic data. These components are then fused using the traditional RGB algorithm. However, the RGB blending algorithm can only fuse three frequency components, which makes it difficult to accurately characterize complex reservoirs. Moreover, it is also difficult to interpret the blending result of white or yellow color. To address these issues, we propose an adaptive Uniform Manifold Approximation and Projection (UMAP) based workflow for multiple-frequency attribute blending and apply it to hydrocarbon reservoir characterization. We first introduce the suggested workflow for multi-frequency attribute fusion analysis. Then, we verify its effectiveness and robustness by applying it to synthetic data. Finally, we apply the suggested workflow to a 3D field data volume and make detailed comparisons with the traditional RGB and PCA based blending techniques for characterizing hydrocarbon reservoirs.