Abstract: Carbonate karst reservoirs of the Lower-Middle Ordovician in Tahe oilfield feature in large burial depth,irregular distribution,complex reservoir types and strong heterogeneity,thus reservoir type identification is always a challenging during oil and gas exploration and development in the area.This paper documented a new impedance inversion method,i.e.waveform-indication-based inversion,based on Bayesian discriminant theory and Markov Chain Monte Carlo sampling algorithm.The method,based on reservoir type identification,seismic reflection characteristics and petrophysical analyses,can be used to optimize sample wells by observing seismic waveform similarities,and establish an initial impedance model by referring to sampling spacing and curve characteristics.The relationship between proposed distribution and prior information is firstly established to achieve efficient sampling of a priori solution space.A Metropolis-Hastings sampling algorithm is then used to sample the posterior probability distribution,so as to obtain the maximum posterior probability solution.The application of the method in identifying the types of the Ordovician karst reservoirs in Tahe Oilfield shows significant improvement of inversion accuracy and a better utilization of horizontal variation of seismic waveforms to reveal effectively the types and spatial and lateral distribution of the reservoirs,and provides fine characterization of the reservoirs.

Key words: waveform-indication-based inversion, Bayesian theory, carbonate reservoir, Ordovician, Tahe oilfield, Tarim Basin