The impedance of sand reservoir will become lower if the pores are filled with natural gas,so the gas reservoir can be detected by using impedance inversion,which is a common used method.However,the impedance of sand reservoir can be influenced by many kinds of factors such as porosity and mineral composition.As a consequence,the impedance of gas sand may be higher than that of high porosity brine sandstone.In this case,mistake would be caused if post-stack impedance is only used to predict gas sand.Elastic impedance is more reliable than acoustic impedance,because gas sand reservoir can induce AVO anomaly.Simultaneously,in order to improve inversion resolution and accuracy,the authors introduce BP (Basis Pursuit) algorithm to complete elastic inversion.This algorithm is used to decompose seismic signal to even and odd wavelet dictionaries and then reflection coefficient can be obtained by the decomposition coefficients.The method proposed by the authors doesn't need initial low frequency model that traditional inversion method SSI (Sparse Spike Inversion) has to know beforehand. In this case,resolution and accuracy can be improved.The application to synthetic data and field data indicates that this method is more accurate than SSI.