The characterisation of karstified carbonate reservoirs is a challenging task due to: (a) Complex fault systems; (b) Amplitude anomalies (karst features) and (c) Strong heterogeneity. To deal with these difficulties, I have developed three techniques targeting at the karstified carbonate reservoirs from Tarim basin:
(a) Fracture detection based on seismic coherence can be affected by dipping reflectors and noise. A robust algorithm is developed, where dip is used to eliminate the impact of dipping reflectors and a pseudo-multichannel appraoch is employed to improve robustness (good performance for S/N ratio greater than 2).
(b) In the presence of karst fractures (strong amplitudes), seismic facies classification based on seismic trace shape may not analyse the weak amplitudes in the background strata. To combat this difficulty, I use the amplitude variation pattern (normalised seismic segment) as the input to avoid the potential impact of strong amplitudes and help to analyse the weak amplitude.
(c) It is difficult to honour the lateral variation due to the strong heterogeneity in the karstified carbonate reservoirs. I adopt a Fourier integral method (FIM) in the stochastic inversion where the lateral variation is honoured based on the seismic trace similarity.
After I have successfully developed and verified these methods listed above, I apply them together to characterise another carbonate reservoir. This case study has less number of existing wells (five). Explicitly, I apply in order: seismic coherence, seismic facies classification, FIM-based stochastic inversion, and the lithofacies classification. I use seismic coherence as prior knowledge in the classification, enabling to honour the localised dolomitization along faults and fractures. Finally, I propose three well locations for potential oil/gas production based on the classified facies and lithofacies probability (over 40%). My study shows that the algorithms can be employed for reservoir characterisation and provide improved results comparing with conventional methods.