An atmospheric surface forcing data set with synoptic temporal resolution is constructed for the U.S. Joint Global Ocean Flux Study (JGOFS) Bermuda Atlantic Time Series (BATS) site for 1988--1992. The forcing data set is based primarily on the 6-hourly European Centre for Medium Range Weather Forecasts (ECMWF) operational analysis, daily cloud fraction and surface insolation estimates from the International Satellite Cloud Climatology Project, and monthly derived satellite precipitation estimates from the microwave sounding unit. Good agreement is found between the ECMWF surface properties (e.g., wind speed, air temperature) and synoptic meteorological data from the Bermuda airport and Comprehensive Ocean-Atmosphere Data Set (COADS) ship reports, though the analysis tends to damp the amplitude of extreme weather events. Monthly air-sea heat and freshwater flux estimates are generally consistent with climatological estimates for the BATS region. The diagnosed net heat and freshwater fluxes from the BATS conductivity-temperature-depth data show significant additional month to month variability that is not related to local atmospheric forcing but appears to arise from mesoscale advection. The surface forcing data set is then coupled to a one-dimensional upper ocean boundary layer model, and the resulting simulations quantitatively reproduce much of the observed behavior of sea surface temperature, heat content, and mixed layer depth for the BATS site for the period October 1988 through September 1992. The induced variability in the ocean model on diurnal and storm timescales is analyzed, and the impact of using the ECMWF analysis data rather than synoptic ship or mooring observations is also examined. The main deficiencies in the simulation are related to the influence of advective events in the BATS record and to possible shifts in the ECMWF model, and preliminary techniques for addressing these problems by incorporating the horizontal advective effects are presented. The difficulties associated with directly verifying local one-dimensional models using coarsely sampled time-series data is also discussed. ¿ American Geophysical Union 1996