The surface concentration of dissolved inorganic carbon (DIC) at the Bermuda Atlantic Time-series Study site (BATS) decreased gradually by ca. 30 mu mol kg(-1) from April to October in 1989. This decrease occurred almost in the absence of measurable nitrate in the mixed layer. Although ancillary data about the C system point to the possible importance of lateral transport, horizontal gradients in surface [DIC] and the mean flow in the area indicate that local effects should prevail in the seasonal drawdown of DIC. On the basis of an one-dimensional model, we hence estimate the mixed layer budget of DIC for this period, from surface [DIC] data, temperature profiles, and concomitant meteorological records. According to model uncertainties, the [DIC] drawdown should be mostly explained (71-93%) by a net community production (NCP) averaging 1.4-2.3 mgC m(-3) d(-1), and to a lesser extent, by outgassing of CO2 to the atmosphere. These losses are partially compensated (<30 %) by mixing with DIC-rich waters of the thermocline. This NCP must be regarded as a lower estimate, since the mean flow from the northeast should bring waters with slightly higher [DIC] to the mixed laver al the BATS site. The model, which is sensitive to short-term variations in atmospheric forcing (<1 day), indicates that this layer has never reached the nitrocline for spring-summer 1989, even as a hurricane passed through the region. Hence, the surface NCP should not have been supported by unsampled, pulse-like supplies of deep nutrients. Wet atmospheric deposition of nitrogen measured concurrently on Bermuda could contribute to the biological N requirement (10-20 %). According to historical estimates, N-2 fixation seems however insufficient to meet the remaining demand. Comparison between NCP and primary production measured in situ suggests that most of photosynthetically fixed C (>50 %) is not respired in the mixed layer.
