Abstract:Hydrocarbon prediction technique is one of people's interests in seismic data processing and interpretation. Having analysed the principles arid problems of CDP stacking and linear convolution model, I deem that existing hydrocarbon prediction techniques based on horizontal stacked (rnigrated) data are only applicable to hydrocarbon prediction in the area where surface arid subsurface conditions are simple, but fails to achieve hydrocarbon prediction in complicated area. In order to improve the reliability arid applicability of hydrocarbon prediction techniques, people must do accurate data processings before stacking, make inversion by using prestacked informations, research into new theoretical model of seismic records, and thoroughly as certain the reflection mechanism of seismic waves from hydrocarbon reservoir.