This study explores the relative impact of three major components of the tropical Pacific Ocean observing system on data assimilation analyses. The extensive observing components that are available during the 2-year period from October 1992 through September 1994 include sea level derived from TOPEX/POSEIDON altimetry, thermistor data from the Tropical Ocean/Global Atmosphere Tropical Atmosphere-Ocean (TOGA TAO) mooring array, and expendable bathythermographs (XBTs) from the Volunteer Observing Ship program. In the first part of this study, methods are introduced to assimilate these data into a primitive equation model using optimal interpolation. The resulting velocity and thermal analyses are then compared with a numerical simulation that has no data assimilation. The most innovative aspect of the assimilation procedure is the development of a statistical model that uses satellite altimetry to update the subsurface thermal field. To determine the impact of individual components of the observing system, three additional experiments are conducted. In each experiment, one component is withheld from the assimilation. The resulting analysis is then compared with the missing data set. Our results show (1) To resolve major features of the seasonal cycle, it is necessary to have either the altimeter or the XBT data. In the absence of XBTs the analysis appears to develop a significant temporal drift in its thermal structure. (2) Intraseasonal variability such as tropical instability waves is best resolved using altimetry because of that data set's superior spatial resolution. (3) Thermistor data from the TOGA TAO mooring array are helpful, but not as crucial, in resolving the seasonal cycle within ¿8¿ latitude as the XBTs. The mooring array is too coarsely spaced to allow this analysis system to resolve tropical instability waves. ¿ American Geophysical Union 1996