AbstractWe investigate mesoscale turbulence (10–1000 km) in the ocean winds over the Kuroshio Extension ( N– N, E– E) using the QuikSCAT data set (November 1999 to October 2009). We calculate the second (Djj) and third‐order structure functions (Djjj) and the spatial variance (Vj) as a function of scale r ( denotes, respectively, the longitudinal (divergent) and transverse (vortical) component). The most interesting results of the analysis follow. Although both and measure the turbulent kinetic energy (TKE), we find that is the more robust measure. The spatial variance density (dVj/dr) has a broad peak near 450 km (close to the midlatitude Rossby radius of deformation). On interannual time scales, TKE correlates well with the El Niño 3.4 index. According to turbulence theory, the kinetic energy cascades downscale (upscale) if (also skewness ) is negative (positive). Our results for the Kuroshio Extension are consistent with a downscale cascade (indicating convergence dominates). Furthermore, classical turbulence theory predicts that and independent of r; however, we find SL varies strongly with r, from −4 at small scales to −0.3 at large scales. This nonclassical behavior implies strong‐scale interaction, which we attribute to the rapid, and sometimes explosive, growth of storms in the region through baroclinic instability. Finally, we find that ST (a measure of cyclonic/anticyclonic asymmetry) is positive (cyclonic) and also varies strongly with r, from 4 at small scales to 0.5 at large scales. New turbulence models are needed to explain these results, and that will benefit Weather Prediction and climate modeling.