Fine-scale characterization of reservoirs is critical for oil and gas exploration and production. Reservoirs in the Z gas field exhibit considerable burial depths and superposed sandstone and mudstone impedance characterized by dark spots, complicating their identification using conventional seismic methods. This study achieved the characterization of deep reservoirs in the Z gas field based on the dominant-channel stacking technique. First of all, this study analyzed the seismic response characteristics of deep reservoirs through forward modeling, positing that poor physical properties are the primary factor contributing to the formation of dark spots. Then, starting with the investigation of the near-well prestack gathers of the target layer, this study examined their amplitude variation with offset (AVO) characteristics and seismic phase stability at different angles, determining the dominant-angle seismic channels that can accurately reflect the phase stability of reservoirs. Finally, this study derived a lithology-sensitive 3D data volume through partial stacking of dominant-angle seismic channels, achieving the characterization of deep reservoirs. The near-trace superimposition (3°~15°) of phase-shift seismic data (-90°) allowed for effectively characterized the H3a reservoir in the Z gas field, providing a geophysical basis for subsequent well deployment. Therefore, the dominant-channel stacking technique can effectively characterize the spatial distribution of deep river-channel reservoirs, guiding the reservoir characterization of similar oil and gas fields.