Using high frequency radar observations of the ocean surface and objective analyses of onshore observations to estimate winds over Monterey Bay, California

Multi-channel radars (MCR) with Doppler capability have proven useful for remotely estimating current profiles in the top few meters of the ocean. We will describe how the same MCR information was used to estimate low-level winds over Monterey Bay, California. The method succeeds because current shear relates to the friction velocity at the water surface, which in turn depends on wind speed, and because differences in radar returns from ocean waves traveling into the wind versus those traveling with it provide information about wind direction. The problem is to extract the information available in the radar returns and turn it into estimates of vector winds. An approach based on partial least squares regression (PLS) proved quite successful at estimating measured winds at a buoy operated by the Monterey Bay Aquarium Research Institute (MBARI). The PLS technique will be described briefly. PLS was applied to 32 parameters measured by two MCRs, each operating at four different frequencies. One of the MCRs is located on the north shore of Monterey Bay, the other on the east shore. PLS is a method that requires “training.” Wind observations at the MBARI buoy were used for this purpose, then the resulting relationships were applied to estimate wind speed and direction from an independent (not used for training) data set collected at the MBARI buoy. The speed and direction biases (estimated minus observed) for the independent data were –0.7 m s–1 and –6.5°, respectively; root-mean-square errors were 1.5 m s–1 and 41°. We will suggest how the PLS wind estimation technique can be made more general. Finally, onshore observations were objectively analyzed, using the Winds on Critical Streamline Surfaces (WOCSS) code which will be discussed briefly. The WOCSS analyses were performed with and without MCR winds as inputs. The results show that the inclusion of radar wind estimates improves agreement with the MBARI observations.