Aspects of the dynamics of internal solitary waves evolving in a three-layer ocean are investigated using a new numerical multilayer model that solves the nonlinear, weakly nonhydrostatic Boussinesq equations and uses high-resolution in situ data. The model applications refer to two different phenomena frequently observed in the real ocean, which can be described using a three-layer model rather than a two-layer model. In the first application the influence of the strength of a shallow seasonal thermocline superimposed on a two-layer permanent stratification on the structure of internal solitary waves is studied. It is found that while for small to medium wave amplitudes a decrease in the strength of the thermocline yields an increase in the simulated wavelengths, for large wave amplitudes this dependence is no longer monotonic. In particular, in the limiting case of a vanishing thermocline, first-mode internal solitary wave solutions of the three-layer numerical model tend to the analytical internal solitary wave solutions of the Miyata equations, a two-layer model, in which the full nonlinearity of the shallow-water theory up to first-order phase dispersion is retained. In the second application that refers particularly to high-resolution observations made north and south of the Strait of Messina in the Eurafrican Mediterranean basin the generation of internal solitary waves by the evolution of surface and subsurface water jets is investigated. The analysis of the in situ data shows in fact that from very energetic surface and subsurface jet-like disturbances subject to strong turbulent mixing internal solitary waves emerge as robust, quasi-nondissipative oceanic features. Idealized flow conditions aimed at approximating possible initial stages of the observed water jets are imposed to force our numerical model. In general, good agreement is found between characteristics of observed and simulated wave fields. Our investigation identifies the observed water jets as peculiar features of the complex ocean dynamics and suggests that layered numerical models can represent helpful tools in understanding fundamental processes inherent in their intricate dynamics. ¿ 2001 American Geophysical Union