A Feasibility Study Of Three Term Avo
Honours Degree, 2002
University of Adelaide
Typical 2-parameter amplitude versus offset (AVO) analysis provides information on contrasts in compressional velocity and either shear velocity or Poisson's ratio across an interface. Whilst these elastic parameters are sensitive to the presence or absence of gas in the pore spaces of a rock, they are ambiguous with respect to gas saturation. A successful 3 parameter analysis would provide density information which has the potential to resolve this ambiguity.
This thesis develops and tests a 3 parameter AVO analysis using the generalized linear inversion (GLI) approach. This problem is constrained to consider the reflection from a single interface, for which the elastic parameters of the top layer (the seal) are known. The inversion determines the three elastic parameters ( , and ) of the underlying reservoir rock. This simulates using AVO to quantify pore fluid change in a reservoir overlain by a laterally consistent seal whose properties are known. Tests showed that as currently implemented, it is necessary for the user to input initial model parameters for the reservoir, larger that the true values in order for the GLI algorithm to converge to a solution.
The major set of tests were performed to investigate the ability of the algorithm to provide accurate results for different maximum angles of incidence at the target interface, and for different levels of random noise in the data. These tests were conducted for representative Class I, II and III AVO models. The results show the importance of long offsets (large angles) for accurate inversions, particularly in the presence of noise. This is because AVO behavior is more diagnostic at larger angles of incidence. The algorithm failed to converge or gave inaccurate results for angles limited to 20 or 30 degrees, particularly when the data was noise contaminated. In contrast, accurate inversions were generally obtained when the maximum angle was increased to 50 or 60 degrees.
Another test to back-calculate gas saturation from the GLI parameters demonstrated the extreme sensitivity of this calculations to small errors in the inversion results. This work was conducted solely on model data, and so results are better than might be expected for real data. Nevertheless, 3 parameter inversion may be feasible for good quality data having significantly large angles of incidence.