Sonar performance prediction relies heavily on acoustic propagation models and environmental representations of the oceanic area in which the sonar is to operate. The performance estimate is derived from a predicted acoustic eld, which is the output of a propagation model. Though well developed nowadays, acoustic propagation modeling is limited in practice by simpli cations in the numerical methods, in the environmental structure to consider (for computational reasons), and even in the knowledge of some environmental properties. This is complicated by the fact that, in sonar performance prediction, the environmental properties need to be predicted for a far future, in the order of hours or days. These limitations imply that the acoustic eld at the output of the acoustic predictor is biased, in current methods. In mathematical terms, the prediction of the acoustic eld can be seen as a model parametrization problem, in which the model is a numerical propagation model, and the parameters are environmental descriptors which, when fed to the propagation model, best model the future acoustic field. Since the 1980's, signi cant research has been done in the development of propagation model parametrization, using techniques of the so-called \acoustic inversion" family. These techniques, having as objective the estimation of environmental properties of an oceanic area, use observed acoustic elds at the area, to be matched with candidate elds corresponding to candidate environmental pictures. At the end, the best acoustic match gives the estimated environment, in other words, the best model parameters to closely reproduce the measured acoustic eld. In the current work, the technique of acoustic inversion is used in the design of an acoustic predictor, together with oceanographic forecasts and measures. Synthetic acoustic data generated with oceanographic measures taken in the MREA'03 sea trial, is used to illustrate the proposed method. The results show that a collection of environments estimated by past acoustic inversions, can ameliorate the acoustic estimates for future time, as compared to a conventional method.