Travel-time-constrained subset selection for sound-speed profile simplification

High-resolution sound-speed profiles (SSPs) improve the fidelity of underwater acoustic ray tracing while increasing the computational cost of layerwise integration and repeated ray-parameter searches. To reduce this computational burden, conventional SSP simplification methods generally construct compact profiles using profile-representation or travel-time-related criteria. However, these methods may retain nodes that are redundant for a prescribed acoustic task or remove geometrically subtle nodes that still affect travel-time predictions. In this study, SSP simplification is formulated as a travel-time-constrained subset-selection (TCSS) problem over the original SSP nodes. The formulation minimizes the number of retained nodes while keeping the maximum travel-time error over a prescribed task-ray set within a specified tolerance. The resulting combinatorial search is performed using a bounded constraint-adapted sequential backward floating selection method. Experiments on 139 Argo-derived SSPs from the South China Sea show that, at a tolerance of 50 μs, the TCSS method retains a median of 12 nodes, compared with 15 for the travel-time-weighted adaptive simplification (TTWAS) method and 27 for the maximum offset of sound velocity (MOV) method. Under equal-cardinality conditions, the TCSS method reduces the mean root-mean-square travel-time error by 68.23% and 88.78% relative to the TTWAS and MOV methods, respectively.