During 1986--1987, an array of horizontal electrometers (which measure the subinertial barotropic velocity) was deployed for 11 months in the central North Pacific to study atmospherically forced motions in a low eddy kinetic energy region. Significant coherence is found to be ubiquitous between the barotropic velocity and one or more surface atmospheric variables over the North Pacific. The maximum squared coherence between currents and wind stress curl at all periods greater than 3 days is typically 0.6, rarely below 0.4, and occasionally above 0.8, and is dominantly nonlocal, although local coherence is seen at some sites, especially at the shorter periods. Both the coherence and the intersite pattern similarity are strongest at the periods of relative maxima in the barotropic current autospectra (e.g., at 6--8, 10--14, and 20--50 days). The locations at which the wind stress curl forces the motions at each site and for each period have been determined. While there is substantial intersite variability, some of the observed patterns of squared coherence over the North Pacific are very similar to those predicted by simple, linear, smooth-bottomed, quasi-geostrophic models of atmospherically forced barotropic currents. The forcing location is occasionally found to be the same spot for all oceanic instruments, especially at the autospectral maxima in barotropic current. More commonly, distict forcing locations are found for each instrument even when closely spaced, suggesting topographic influence.

In other cases, multiple forcing locations are implicated for the motions at a single station. All of this site- and period-dependent variability of the forcing location is incompatible with the simple models and is due either to the known spatial inhomogeneity of the curl forcing field or to topographic effects. The importance of the former is emphasized using simple arguments based on analytic transfer functions between the atmospheric and oceanic variables and on observed wavenumber spectra for wind stress curl. Quantitative assessment of topographic effects must await suitable numerical models. In addition, maximum likelihood wavenumber spectra of the barotropic currents have been estimated in those bands with an adequate number of coherent sensors. The wavenumbers at the spectral peaks are compatible with the dispersion relation for Rossby waves modified by the observed, large-scale, linearly sloping topography which shoals to the northeast. The observed group velocities are in general quite consistent with the inferred forcing locations based on the ocean-atmosphere coherences. ¿ American Geophysical Union 1992