Sand-rich deposits in the submarine canyon system in the central Gulf Coast region of Texas are considered as high exploration potential reservoir bodies. However, the multiphase regression and transgression during the early Eocene complicate the structure position and the sand body distribution. Seismic attributes have been proven to be sensitive to seismic amplitude for detecting channels and canyons for decades. Specifically, the variance attribute is useful in distinguishing stratigraphic features, the root-mean-square (RMS) amplitude attribute is a good indicator of amplitude anomalies, and spectral decomposition is specialized in revealing the variations of layer thickness. In this study, we combine the variance and RMS amplitude attributes to detect the edges of the canyon and perform spectral decomposition to analyze the varying thickness of the sand bodies. Identification of the canyon system can provide significant constraint on predicting the distribution of the reservoir sand, and consequently improve the evaluation of production potential.