Different Approaches to Shear Wave Prediction

O.H. Afif¹, M. Ahmed¹, and H.H. Soepriatna¹

Abstract

Shear velocity is a key input for rock physics analyses supporting AVO inversion workflows. There are many cases where measured shear logs are absent or unreliable. Therefore, it is necessary to predict them. The main goal of this study is to do shear log predictions and examine the limitations of different estimation techniques. There are many techniques to predict the missing shear velocity. We examine three of these methods within our study area. Empirical models are the first of these techniques. Many empirical relations use different number of petrophysical logs to estimate shear velocity. Castagna (1985), Han (1986), Tosaya (1992) and others have published on different aspects of deriving accurate shear velocities. The second technique employs theoretical rock physics models, which is a mechanical approximation of the elastic, viscoelastic, or poroelastic properties of rocks (Avseth, 2005).The third technique uses an artificial neural network approach that develop efficient relationships between the input data and the target predicted data. Petrophysical logs and key geophysical logs, such as compressional sonic and bulk density, were properly preconditioned and rigorously edited within several intervals of interest. In an attempt to predict shear velocity, we tested many empirical relationships in our study area and concluded that their limitations are due to the underlying assumptions made by these relationships. We conclude, consequently, that the tested empirical relationships are not necessarily applicable to our reservoir. We found in our study area that the empirical relationships that use compressional velocity to estimate shear velocity are better than other methods that use porosity or clay content. Gassmann (1951) was used to remove the fluid effect in order to meet the rock physics models assumptions. From this we found that the rock physics models are very sensitive to petrophysical inputs: any discrepancies or missing petrophysical information invalidates the rock modeling results. We used the neural network trained with key logs to predict the missing shear wave velocity. Results from the three estimation techniques were evaluated in the elastic domain to analyze the dynamics of the predicted versus measured shear. We conclude that both the petrophysically-constrained neural networks and the rock physics models can be used successfully to derive shear velocities in our study area.

AAPG Search and Discovery Article #90188 ©GEO-2014, 11th Middle East Geosciences Conference and Exhibition, 10-12 March 2014, Manama, Bahrain