Abstract:The horizontal prediction of thickness and porosity of a sand member in a region was made by applying multivariate statistical method to seismic attribute,drilling and logging data. The multivariate statistical method mainly involves significant figure calculation,stepped regressive analysis and Krig interpolation. Significant figure determines which seismic attribute may be applied to reservoir prediction,and useful seismic attribute can be thus chosen properly. The stepped regressive analysis uses the chosen seismic attribute to estimate reservoir parameter,by which efficiency and accuracy of prediction can be improved obviously. However,the estimation result slightly deviates from real result at the borehole point where model is constructed. The deviation will be reduced further by doing Krig interpolation according to the structural property of residual deviation. Its computation result of a land-facies thin interbed shows that the accuracy meets the requirements of exploration or development phase.