Several time series of wind components and X-band Doppler radar signals, gathered concurrently over a approximately 0.01 km(2) area of the ocean surface, were examined for evidence of a low-dimensional dynamical attractor with the Grassburger-Procaccia algorithm. Only the vertically polarized radar reflectivity and the horizontal surface wind speed time series suggested the presence of such an attractor. The correlation dimension for these two observables appeared to be nearly the same. This suggested a working hypothesis that the dynamical behavior of both the vertically polarized radar reflectivity and the horizontal surface winds are controlled by a single low-dimensional dynamical system. The hypothesis was further examined by predicting winds from radar reflectivity, using a neural network deterministic model, and comparing the prediction performance with that of the SEASAT statistical algorithm for retrieving surface winds from radar backscatter. It was found that the deterministic model did, in fact, achieve a higher prediction correlation coefficient for a limited time period. (c) 1995 American Institute of Physics.