Abstract: To address the challenges of rapid lateral reservoir variations and difficulty in identifying lithologic trap boundaries in fluvial-tidal coupled environments, this paper proposes a reservoir prediction method based on ocean bottom node (OBN) seismic data. Using common reflection points as the basic computational units within sedimentary facies zones, this method overcomes the scale limitations of traditional channel analysis by automatically extracting dominant azimuths to optimize azimuth division. On this basis, azimuth-specific inversion of reservoir-sensitive parameters is performed to obtain multi-azimuth attribute volumes, enabling accurate determination of dominant anisotropic orientations. To further integrate the high signal-to-noise ratio of full-azimuth data with the strong anisotropic information from divided azimuths, a global optimization objective function for multi-azimuth attribute fusion is constructed to adaptively determine weighting parameters. Application results demonstrate that the proposed method significantly improves the delineation accuracy of complex sand body boundaries, with fusion results showing strong agreement with sedimentary facies distribution. This provides a reliable basis for the exploration and development of complex tide-dominated reservoirs and offers valuable insights for reservoir prediction using wide-azimuth seismic data.