Vector wind field measurements using HF radar

Jessica Drake

Co-Authors:
John Vesecky & Kenneth Laws
EE Dept., Univ. of California at Santa Cruz, Santa Cruz CA 95064
Calvin Teague & Frank Ludwig
STAR Lab.
EE Dept., Stanford Univ., Stanford, CA 94305
Jeff Paduan
Dept. of Oceanography
Naval Postgraduate School, Monterey, CA 93943

It is well known that HF radars are capable of measuring wind direction by using the relative strength of the echoes from the approaching and receding ocean waves at the Bragg resonant wavelengths. Here we examine the ability of multifrequency HF radar to measure wind speed as well as direction. In this study we use data collected over Monterey Bay, California in December of 2000. At that time the M1 buoy (deployed by the Monterey Bay Aquarium Research Institute, MBARI) was in the radar’s observational area, near the Bay mouth, and measured wind speed and direction. Two multifrequency coastal radars (MCR’s) near Santa Cruz and Moss Landing, California operated at 4.8, 6.8, 13.4 and 21.8 MHz, measuring currents at effective depths of 2.5, 1.8, 0.9 and 0.6 m respectively. Using the method of partial least squares we developed an algorithm for estimating the surface wind vector from multifrequency HF radar data. For inputs this method uses the relative echo strengths of the approaching and receding Bragg lines as well as the near surface currents estimated for the four effective depths mentioned above. Partial least squares is a predictive technique based on relationships estimated from a training data set within which both inputs and outputs are known. We use the M1 buoy winds as output ‘truth’. Our work indicates that the method produces excellent results. The U and V wind components were estimated with a standard error of prediction of a little over 1 m/s, biases of less than 0.2 m/s and R2 in the range 0.65 to 0.94. An investigation of the weights in the partial least squares algorithm indicates that the relative echo strength in the Bragg lines, near surface currents and near surface current shear are important in determining the wind estimates. We think that this method will find useful application in measuring the detailed structure of the wind field in coastal regions on a few kilometer size scale.