Fluid Prediction from Time-lapse Seismic AVO Data

Reliable reservoir characterization of both the porosity/permeability and the fluid distribution is important for reservoir management. Geoscientific experience, seismic data with good coverage and logs along a small number of well traces provide the basis for this characterization. The time-static porosity/permeability distribution is challenging to assess, while the time-dynamic fluid distribution is even more challenging to monitor. Reliable characterization of the dynamic fluid filling is crucial for reservoir engineering management including the design of efficient infill well drilling programs. Our study is focused on prediction of the fluid dynamics based on time-lapse seismic AVO data. We apply spatial Bayesian inversion methodology, necessitating a prior model on the reservoir characteristics. This is challenging because the saturation is bimodal. We present a solution using a selection Gaussian prior model and a Gauss-linear relationship between the reservoir characteristics and the seismic responses. The methodology is tested on seismic data generated from well observations from Kneler in the Alvheim oil and gas field. The results are encouraging, preserving the bimodality of the saturation even in the presence of considerable error.