Within the deep-water ocean (depth >1000 m), low frequencies (1-300 Hz) acoustic waves are characterized by very long range propagation. In this context, the propagation is largely impacted by the oceanic environment. Thus, localization methods based on classical plane wave models do not perform well. Matched field processing has been proposed to include better environmental models, but it is known to perform poorly as soon as the environment is not perfectly known. Given that the ocean is a dynamic system, it changes quickly over time and space making this method inapplicable in an operational context. To circumvent this issue, a better way is to consider methods that do not require detailed knowledge about the environment.In this manuscript I consider a quantity called the Waveguide Invariant (WI) which is known to be robust to small environment variations. In particular, I investigate two phenomenons responsible for its dependence to the source-receiver configuration: the dominance of the acoustic field by groups of modes and the frequency dependence of the Eigenmodes. Using a ray-mode approach, these two features are integrated in a WI derivation which provides a thorough way to predict and understand the striation patterns in deep-water context. Then, using this underlying physics driving the propagation along with the concept of mode trapping, a depth localization method is proposed. The input data for the algorithm is a range-frequency intensity, as measured on a horizontal line array. This idea is explored and extended to propose a source depth discrimination which is performed as a binary classification problem.