Abstract
In reservoir characterization, 3-D seismic data have been used as soft data for geostatistical analysis. The reason is that the seismic data is acquired uniformly in the reservoir area. On the contrary, the well data are located limitedly. The one advantage of using seismic data is that the information of spatial variation can be easily derived. The second one is that we can reduce the uncertainty in geostatistical analysis, conditioned by the cross relationship between the reservoir property and the seismic data.
The estimation variance in cokriging is less than that in kriging because of that cross relationship. In the case study using two oil fields, the estimation variances in kriging and in cokriging were compared with the errors between the actual data and the estimation by cross validation tests. In general, the errors evaluated in cokriging were less than that in kriging. But it was found that the large dispersion between the actual data and the estimation can occur, in the case of hiding the well data which has an effect on priori statistics, and in the case of using only small number of well data. In that case, the estimation variances in cokriging had been underestimated. When only poor sample data are available, the reduction of uncertainty in cokriging seems to be unreliable.
