Discussion on variable-velocity depth mapping methods for deep-ultra-deep complex structures

High-precision depth conversion is particularly challenging along the southern margin of the Junggar Basin, where seismic velocity fields exhibit severe lateral variations in deep-ultra-deep, structurally complex areas. Taking the low-relief Dunan anticline and the ultra-deep Hutubi anticline as examples, this paper explores velocity modeling strategies tailored to different geological settings and data conditions. Three practical methods were systematically studied and applied: 1) a joint well-velocity method combining average and interval velocities from wells; 2) a 3D layer-controlled method integrating seismic stacking velocities with well data; 3) a multi-source 3D layer-controlled method incorporating stacking velocities, well interval velocities, and uphole survey data. The results show that for deep low-relief structures with only a single well (e.g., Dunan), method 1), aided by pseudo-wells and trend-surface constraints, effectively controls the velocity trend, yielding a structural error of less than 5%. Method 2), which extracts interval velocities from stacking data and populates a 3D layer-controlled model based on their lateral trends, maintains the same error level (< 5%) while significantly improving modeling efficiency and lateral control. In ultra-deep zones with near-surface high-velocity conglomerate bodies (e.g., Hutubi), method 3), through detailed 3D layer-controlled modeling that incorporates these anomalies, substantially enhances velocity accuracy, resulting in depth maps with errors of less than 1% compared to well tops. This study emphasizes that no single velocity-modeling technique is universally optimal for deep-ultra-deep exploration. The key lies in selecting and blending appropriate methods based on the specific geological complexity, structural style, data quality, and control-point density of the target area, and in iteratively constraining and quality-controlling the model using multi-source information such as seismic, well, and VSP data.