For a 10-day period during September 1990 the R/V Franklin worked around a drifting buoy drogued at 20-m depth in the Bismarck Sea near 4¿S, 149¿E. Continuous measurements were made of the air-sea fluxes of radiation and sensible and latent heat, and a conductivity/temperature/depth cast to 400 m was made about every 6 hours. The aim was to close the heat budget of a sample volume of the surface mixed layer to within 10 W m-2, in preparation for our participation in the 1992--1993 Tropical Ocean and Global Atmosphere-Coupled Ocean-Atmosphere Response Experiment (TOGA-COARE). Temperature and salinity between the surface and 30-m depth were quite uniform, but below 30 m, variability was observed which suggested the possible intrusion of horizontal and vertical advection of heat. Heat content was analyzed for depths of 40 m and 20 m; bulk Richardson numbers generally greater than 0.8 and 0.4, respectively, in the two cases indicated that diapycnal mixing through the bottom of the 40-m volume could be neglected at 40 m and possibly at 20 m (Peters et al., 1988; Godfrey and Lindstrom, 1989). An eddy diffusivity for salt at 20 m was obtained to account for the steady decrease of observed freshwater content in the top 20 m over that expected from the surface flux.

Using this diffusivity, the turbulent heat flux through 20 m was of order 6 W m-2, supporting the view that vertical mixing of heat was small even at this depth. Then, neglecting advection and vertical mixing, the heat budget closure to 40-m depth was satisfied to about 25 W m-2 on average over the period, but both integrated heat and freshwater time series were ''noisy'' because of variability below 30 m. Limited to 20-m depth, the average difference between incident energy and heat content was reduced to about 12 W m-2, with close agreement over the diurnal cycle. The model for air-sea exchange of sensible and latent heat by Liu et al. (1979) is verified at low wind speeds, although it may overestimate slightly for winds over the range 3--6 m s-1. This study has identified several possible sources of measurement, parameterization, and sampling error in determining the net heat flux into the ocean: however, with good sampling for the advective components, errors in heat budgets should be within the 10 W m-2 accuracy aimed for during COARE. ¿ American Geophysical Union 1993