A hydrographic dataset based on data from the SEQUAL/FOCAL experiment is used to determine the errors of a numerical simulation of the complete temperature and velocity fields of the tropical Atlantic during the two-year period 1983–84. To improve the accuracy of the analysis we develop an application of four-dimensional data assimilation. In this analysis the thermal fields of the model are updated once a month using sea surface temperature measurements and observed temperature profiles. Much of the paper describes comparisons between differing analyses using data assimilation and the numerical simulation, and verification of these with temperature and velocity data from moored instruments. Assimilation of the temperature observations improves the accuracy of the temperature analysis. The amplitude of seasonal changes in the meridional thermal gradient is doubled at 38°W, bringing the analysis closer to the observed thermal gradient. At 28°W the improved is less dramatic. The zonal thermal gradient at the equator is increased, but the month-to-month variability also significantly. Comparisons are made with temperature and velocity measurements at midbasin mooring sites. Here assimilation sharpens and lowers the thermocline and reduces long-term trends in the thermal field. Assimilation also improves some features of the velocity field such as the depth and eastward penetration of the undercurrent core and the strength of the North Equatorial Countercurrent.