An Ocean General Circulation Model (OGCM) of the tropical Pacific in which combined TOPEX/Poseidon and ERS sea level anomalies are assimilated over January 1994 through July 1999, is used to investigate equatorial wave characteristics during the intense 1997--1999 El Ni¿o-La Ni¿a event. Near the equator, the linear vertical modes are estimated at each grid point of the OGCM simulation with and without assimilation. Consistently with an increase of the vertical gradient within the thermocline and a rise of the thermocline depth in the eastern basin, the assimilation results in an increased contribution of the higher-order baroclinic modes in the eastern basin and a decreased contribution of the first baroclinic mode in the western Pacific for the zonal current variability. For pressure, the first baroclinic mode contribution is reduced whereas the higher-order modes contribution is weakly impacted. Kelvin and first-meridional Rossby waves are then derived for the first two more energetic baroclinic modes in the simulation with assimilation. Kelvin waves of both modes constructively contribute to the strong warming observed in 1997, with the first (second) baroclinic mode being more energetic than the second (first) baroclinic mode in the early (mature) stage of the warming. Kelvin waves of both modes reflect as first meridional Rossby waves at the eastern boundary (reflection efficiency of ~95%) and contribute to push back the warm pool westward. The reversal of the warming is apparently initiated by the second baroclinic mode contribution which controls the position of the 28¿C isotherm in the surface layer in the far eastern Pacific from January 1998. At the western boundary, reflections of Rossby waves take place for both modes with an estimated total efficiency of ~50% at 165¿E. This suggests that, in our model, the delayed oscillator theory is not applicable for explaining the reversal from warm to cold conditions in 1998 while the zonal advective feedback was at work. More generally, the study suggests that it is necessary to take into account the vertical structure of the ocean when interpreting altimetric data, which can be done through an assimilation experiment.