Addressing non-linearity in tomography using ray theory

Ocean acoustic tomography has been formulated as a linearized inversion of a forward model based on ray tracing. In this formulation, we are solving for refinements of a zero-order sound speed model that reproduce measured travel times, travel times that are also being modeled by the ray tracer. The linear approximation is based on calculating travel time perturbations by integrating the sound speed refinements over the ray paths, assuming the ray paths themselves are not impacted by the sound speed refinements. Clearly, the offset between the presumed (zero-order) sound speed and the actual sound speed may be so great that the ray paths need to be modified, not just the sound speeds being integrated along these paths to reproduce the travel time perturbations. There has been a lot of work on using full wave propagation models to predict the sensitivity of travel times to sound speed perturbation throughout the medium. We will present the use of ray theory, leveraging dynamic ray tracing to gain practical insights into how to alert a user and to overcome non-linearities that arise in tomography.