Uncertainty modeling for ocean tomography systems

In ocean tomography acoustical means are used to infer characteristic parameters of the underwater medium such as temperature profile and current velocity. These techniques are based on the functional dependency between the propagation and the parameters that must be estimated. Accuracy of this modelisation is critical to the performance: a mismatched model may preclude correct inference, either by imposing biases or by affecting the variance of the estimates. The authors apply to this problem a global performance prediction tool showing its ability to predict system's sensitivity to modeling errors.<
>