Geostatistical inversion (GI) combines the advantages of stochastic modeling and seismic inversion, integrates geological, logging and seismic data, and delineates the reservoir more preciously. Constrained sparse spike inversion is conducted to predict the general distribution of reservoirs before GI, and the results including wavelet and 3D acoustic impedance (AI) can be used later. The inversion uses both well-log and seismic data. The inter-well AI is created by stochastic simulation, and the subsequent work includes (1) simulating an AI at the interwell point, followed by computing reflectivity series, (2) convolving it with the wavelet selected to simulate a local trace, and (3) implementing iterations to make the simulated and seismic trace reach a desired fit. The results are of multi-equiprobable 3D AI volume realization, which can be used for uncertainty evaluation. The inversion integrates the high resolution of logging and lateral directions of the seismic data. The results fit the geostatistical features of the input data, and are constrained by structural and stratigraphic earth models.