Atributos AVO/AVA de alta resolución utilizando very fast simulated anneling

we present a new inversion method to obtain sparse-spike avo/ava attributes from prestack seismic data. the proposed method aims to find the smallest number of reflectors that, when convolved with the source wavelet, fit the data. this method is an extension to prestack data of an earlier work on sparse-spike deconvolution that was applied to poststack data. due to the high nonlinearity of the inverse problem, which includes the determination of the time location of a given number of reflectors, we use the global optimization algorithm known as very fast simulated annealing (vfsa). unlike other inversion methods that look for sparse solutions in which the number of unknowns is equal to the number of samples of the seismic data, in the proposed strategy the number of unknowns is much smaller. as a result, the matrices used during the inversion are small and the process is relatively cheap in terms of computational cost, despite the fact that the inversion is carried out using simulated annealing. the technique can also be used to determine additional unknowns such as a constant phase rotation of the wavelet (to calibrate an initial estimate) or its central frequency (to compensate for attenuation effects). one advantage of the method is that the uncertainty of the solutions can be estimated stochastically, taking advantage of the large number of solutions that are tested during the inversion process. the high resolution avo/ava attributes obtained after the inversion include both the intercept and the gradient, i.e. the coefficients of the two-terms shuey's approximation of the zoeppritz equations, which are used to model the variation of the reflection coefficient with the incidence angle. however, the incorporation of other approximations is immediate. results using 1d and 2d synthetic data show that the proposed method is robust under noisy conditions, even in the case where the number of reflectors is not known a priori and the utilized wavelet is inaccurate. it