Decameter wavelength (3 to 30 MHz.) radars operating in ground wave mode are capable of mapping surface currents, waves and winds over relatively large coastal ocean areas (≈ a few thousand km2), Since only radial current speed is measured at a single site, it is necessary to use two or more sites to construct vector current maps, such as Fig. 1 below. Several different HF radar systems are often in operation, using different radar system concepts, processing methods and radar frequencies. Users of the data products of HF radars need to know the likely uncertainties and/or biases caused by the use of radars at different operating frequencies and how to merge the data from these different systems into a single current map. Monterey Bay provides a useful test bed for developing algorithms and tools for processing and merging HF radar observations from different systems, namely SeaSonde units from Codar Ocean Sensors Ltd. and multifrequency coastal radars (MCR's) developed by a consortium based at the University of Michigan. During the year 2000 two MCR's operating at 4.8, 6.8, 13.4 and 21.8 MHz were sited at Moss Landing and Santa Cruz, California, collecting current maps at hourly intervals. At the same time three SeaSonde units were operating from the aforementioned two sites and from Pt. Pinos as well. This suite of radars provides excellent coverage of Monterey Bay, but operates on frequencies covering the 4.8 to 25 MHz range.

In this paper we use the current maps generated by the MCR units to investigate the hypothesis that vertical current shear generated by wind stress produces a bias between currents measured with different frequency HF radars. First, we develop a simple theory for the aforementioned bias using standard, law of the wall, boundary layer theory. This is follow by an empirical demonstration of the bias and a comparison with the wind stress hypothesis. Finally we discuss briefly some methods for compensating for this bias. We argue that these methods forma basis for more effectively merging radar data from different systems than simply ignoring operating frequency.