There is a growing concern in the literature about the effects of low frequency sounds (LFS) on marine mammals. A primary way to assess these effects on marine mammals involves the study of disturbance reactions. Detailed research of disturbance reactions of submerged marine mammals requires 3- dimensional localization and tracking of the animals. Animals such as sperm whales Physeter macrocephalus are localized passively with the use of travel time differences (TTD) of their vocalizations received by multiple hydrophones at known positions. Classically, straight-line paths of sound propagation between source and receiver are used to calculate source position. A more accurate calculation of source position involves naturally occurring non- constant sound speeds. This gives rise to arced paths of sound propagation between source and receiver. An algorithm is used to recursively pinpoint source position in a medium with a non-constant sound speed. 5 hydrophone array configurations are tested, each with 30 randomly generated source positions. Average errors of the 150 source position calculations (x, y, z) are (±1.58m, ±1.70m, ±10.44m) for the straight line, and ±0.76m, ±0.87m, ±1.10m) for the algorithm. On average, the algorithm improves the source depth calculation by an order of magnitude.