Carbonates of Maokou formation in the Jiulongshan area, western Sichuan basin, were subject to long-term weathering, denudation and dissolution, leading to karst developed. As an important lower Permian gas producer, Maokou formation has great potential for natural gas exploration. In order to improve the the prediction accuracy of karst reservoir, the distribution of karst body is predicted qualitatively by analyzing the seismic reflection structure of the overlying strata of karst body, studying the seismic facies of the 2nd Maokou member(P1m²), rebuilding the paleogeomorphology, analyzing the characteristics of the seismic attributes of karst body and so on. Gamma inversion was used to predict the biological beach thickness distribution of P1m² member, and on this basis, multi-attributes inversion was used to quantitatively predict the karst reservoir. The result shows that the prediction method combined with qualitative and quantitative prediction is more effective and reliable for the characterization of karst reservoir, thus providing an important basis for the development of Maokou formation in Jiulongshan area.