An adaptive multi-point geostatistical modeling and inversion method with relative geologic age constraint

The low-frequency model can reveal trends of subsurface media and be an initial value for seismic inversion, which can effectively reduce the multiplicity of solutions and uncertainty of seismic inversion. When the initial models deviate from the true subsurface situation, it will bias inversion results. In order to obtain high-precision initial models, we propose an adaptive and inversion method with relative geologic time/age (RGT) constraint based on FILTERSIM algorithm, which is one of multi-point geostatistical (MPG) modeling methods. Firstly, the principal component analysis (PCA) method is used to replace the predetermined filters using conventional FILTERSIM algorithm, it can adaptively extract effective features from seismic data according to the complexity of underground structures. And, the accuracy of modeling results is improved while removing the limitation on the number of filters. Secondly, the RGT is introduced as a spatial constraint to guide interpolation and extrapolation of well-log data, it can effectively solve the problem of excessive smoothing in interpolation results of traditional methods far away from well-log data, as a result, the modeling results have better spatial continuity. The test results using both synthetic and real seismic data show that high-accuracy modeling and inversion results of elastic parameters are obtained through our proposed method.