How to eliminate mistiness and improve the precision of seismic data interpretation is an everlasting topic in reservoir prediction. However，it is rather difficult to solve the problem due to the complexity of actual geological and geophysical conditions and low quality of seism ic data. Recent study shows that integration of information from different sources is an effective way to deal with the issue. Guided by this understanding，we tried to combine artificial neural network technologies and geostatistics in reservoir prediction of the D gas field. By doing so，we realized the integration of multiple seism ic attributes with logging data，and effectively reduced the mistiness and improved the precision of prediction.