Abstract: The prediction of fractures in the Mesozoic granite buried hills in the Qiongdongnan Basin in the northern South China Sea is difficult, and the lack of wide azimuth pre-stack seismic data results in the inability to carry out AVAz inversion. The low seismic signal-to-noise ratio in the target layer of buried hills results in poor prediction of fractres using conventional seismic attributes (coherence, curvature) analysis. Rock physics analysis of the granite buried hill in the study area indicates that the heavily and moderately weathered zones of the buried hill reservoir exhibit low S-wave impedance characteristics, while the interior fracture zones of the buried hill demonstrate high S-wave impedance characteristics. Based on the stacked seismic data after gather optimization, curvelet transform processing, ant-tracking and maximum likelihood attribute analyses were conducted; shear-wave impedance inversion were performed which yield the relative shear-wave impedance results and the analysis of buried hill reservoir were completed. Research has shown that the ant-tracking attribute and maximum likelihood attribute after curvelet transform can characterize large and medium scale fractures and internal details, and the characterization effect is better than seismic curvature attribute. The shear wave impedance after curvelet transform can reliably characterize the weathered zone reservoir of granite buried hills. This article integrates the ant-tracking attribute, maximum likelihood attribute, and S-wave impedance results of the curvelet transform to establish the development model of Mesozoic granite buried hills in the study area. The prediction results of the buried hill reservoir have been confirmed by drilling, achieving good prediction results. The above method is not only applicable to the prediction of granite buried hill fractures and weathering zones, but also provides reference for the prediction of metamorphic rock buried hill reservoirs, and has good applicability in similar regions.