A comparison between local and global inversion of poststack seismic data to estimate acoustic impedance

Abstract

Changes in the elastic impedance are an important way to consider the spatial variations in rock properties and seismic reflectivity. Usually, impedance is recovered using inversion methods of which there is a wide variety. With local inversion methods, the solution is usually strongly dependent on the initial model; the inversion’s cost function is prone to being trapped in local minimum. Local methods are advantageous in that they are computationally efficient, but they require accurate constraints as input. In order to overcome this disadvantage, we utilize instead a simulated annealing method to invert for the elastic impedance. One benefit of using this type of global inversion, is that the conditions on the initial input model is less severe, that is, only the low frequency trend of impedance with depth or time and more reasonable bounds are needed on the input parameters in order to reduce the nonuniqueness of the solution. Further, the cost function converges more confidently to the global minimum. In contrast with least squares method, computation by using simulated annealing is time consuming. However, better impedance is inverted by using simulated annealing and there is no need to input accurate constraints. However, local methods are less time consuming and expensive. Here, we use a least squares method to invert for impedance and we compare these results to those obtained using simulated annealing.