Seismic horizon picking via marching semblance dynamic time warping
Hao Wu, Weimin Wang, Sandong Zhou · Geophysics · 2025
ABSTRACT Seismic horizon interpretation is critical for reservoir modeling, but it remains labor intensive and subjective when performed manually. Although automated approaches can be efficient, they struggle with complex geologic features (e.g., faults and unconformities) and noisy data. We introduce marching semblance dynamic time warping (DTW), a robust algorithm that enhances horizon tracking by integrating seismic semblance with DTW in a novel framework. Traditional DTW-based methods often fail to maintain lateral continuity and exhibit poor noise robustness, particularly in structurally intricate zones. Our approach replaces DTW’s conventional Euclidean distance with seismic semblance to improve noise resilience and uses a marching mechanism that propagates optimal alignment paths across adjacent traces, ensuring coherent horizon extraction even across discontinuities. The method reduces computational overhead through adaptive windowing while preserving geologic consistency. Tests on synthetic and field data sets demonstrate superior performance over conventional DTW, achieving higher accuracy in faulted regions and improved computational efficiency. This innovation enables reliable large-scale horizon interpretation with minimal user intervention, advancing the feasibility of automated seismic analysis in complex subsurface environments.