Abstract: In the Junlian Region of southern Sichuan with large surface morphology and tectonic fluctuations, the conventional reservoir property modeling method has been used to predict the three-dimensional gas content of coalbed methane. In order to improve the accuracy of 3D model for gas content, the grey correlation analysis and multiple regression analysis were combined to build a gas content prediction model, considering the actual buried depth of each grid, a three-dimensional gas content modeling method suitable for the region was proposed. The results show that the correlation factors of gas content in Junian Region of southern Sichuan were sorted in descending order by relevancy: water, maximum vitrinite reflectivity, buried depth, ash and fixed carbon. The co-linearity test shows that gas content prediction model established by water, maximum reflectivity, buried depth and ash is rather accurate. The fitting correlation coefficient R2=0.701 is significant at 99% significance level. This three-dimensional gas content calculation model took into account the surface morphology and underground structure. The influence of the buried depth was closer to the actual situation. The calculated gas content of the larger deep-buried area was higher than the value from the conventional property modeling method, and the improved gas content modeling result was closer to the actual production situation than the conventional modeling result.