A Tool to Estimate the Sensitivity of Underwater Acoustic Transmission Loss to Environmental Uncertainty

Inputs to mathematical models can be very uncertain because of incomplete knowledge. These input uncertainties propagate to output uncertainties, and an estimation of these output uncertainties, and of the importance of the individual inputs to these output uncertainties, would be desirable. These estimates can be calculated via Uncertainty Analysis (UA) and Sensitivity Analysis (SA) respectively. A software library, Algorithms for Stochastic Sensitivity Analysis (ASSA) has recently been written (Jansen 2005) to carry out these analyses. ASSA estimates the total variance (VTOT) for uncertainty, and for each input parameter, the “top marginal variance” (TMV) for its sensitivity the reduction in variance that would occur if the input were perfectly known. We adapt the ASSA methodology to underwater acoustic propagation. Our tool uses efficient uncertainty models for sound speed profile (SSP) and bottom loss (BL) to compute an ensemble of SSPs and BL curves. A propagation model is then used to compute an ensemble of transmission losses (TLs) from which VTOT and the TMVs are used to estimate TL uncertainty and sensitivity respectively. Example results for a real ocean environment are presented.