Yearlong measurements of current and temperature are used to examine the characteristics of eddy heat flux and its distribution in Drake Passage. This seemingly straightforward calculation can produce misleading results when measurement errors are not corrected and when careful consideration is not given to the nature of the time series under study and the coordinate system used. When moorings blow over during periods of high current speeds, eddy heat fluxes produced from the resulting records may be overestimated by as much as 20% if not corrected by using concomitant pressure time series and vertical temperature gradients. Although short-time-scale processes (such as tides, internal waves, and inertial oscillations) are routinely eliminated from heat flux estimates by low-pass filtering, low-frequency contamination of eddy heat fluxes is not usually considered. Long-period events can impose unwanted, dominating cross correlation (eddy heat flux) between fluctuation temperature and velocity. In Drake Passage, low-frequency variability in some current records is associated with sporadic lateral shifts of the fronts within the Antarctic Circumpolar Current. Defining an effective eddy time scale (40 hours to 90 days) and band-passing the current records before calculating eddy heat flux results in values that are more homogeneous (consistent with one another) in direction and have higher statistical significance. Finally, we consider the coordinate system used for the eddy heat flux calculations relative to the dyanamic processes under study. For the Circumpolar Current we define poleward eddy heat flux as that component perpendicular to the axis of the current that forms a continuous, but distorted, band encircling Antarctica. Since the current direction varies in time, across-stream eddy heat flux is calculated relative to the 90-day low-pass direction. The resulting heat fluxes from all deep instruments are small (about 1 kW/m2), whereas those at depths less than 1000 m in the northern passage are an order of magnitude larger.

The average across-stream flux for all available instruments is 3.7 kW/m2 and directed to the right (poleward) of the 90-day low-pass current. Cospectra of fluctuation temperature and velocity show the heat flux is separated, on the average, into two frequency bands with corresponding periods between 100 and 40 days and 16 and 10 days. This pattern is strongly influenced by the records from shallow instruments and those in the northern passage and compares favorably with the cospectrum from a 5-year bridged record from the central passage. Comparison with previous estimates in this region shows considerable differences, partly as a result of the use of different techniques. However, analysis of five 1-year records from a single location shows the interannual variability is large,which urges caution when interpreting the significance of isolated short-term estimates of eddy heat flux.