Abstract: Seismic inversion can reflect the morphology of subsurface formation and facies information more intuitively. Seismic impedance inversion can directly produce elastic information of underground media, providing reliable information for subsequent coal identification, water channel prediction and collapse column distinguishing. A Bayesian-based inversion of reflectivity method was developed. It can consider the various propagation effects of seismic waves and solve the objective function nonlinearly, which can more accurately calculate the impedance and improve the inversion resolution. The method is applied to the actual data from Buertai coal mine and satisfactory results are output, which effectively verifies the feasibility and effectiveness of the new method. Compared with traditional inversion method, the resolution and accuracy are improved. It can effectively identify the distribution of thin coal and deep coal. It can provide valuable information for identifying the coal and the collapse column by using seismic inversion technology and predicting the water distribution in the roof and the floor.