A new method for estimating high-frequency radar error using data from Central San Francisco Bay

Abstract

This study offers a new method for estimating High-Frequency (HF) radar surface current velocity error in data comparisons with other types of instrumentation. A new method is needed in order to remove the zero-mean random spatial and temporal fluctuations present in surface-current measurements from all sensors. Conventional methods for calculating radar error when comparing with another instrument have included their root mean square differences and scatter plots that provide correlation coefficient and slope/intercept of the regression line. It seems that a meaningful estimate of radar error should attempt to remove both sensors’ zero mean random fluctuations, inasmuch as possible. We offer and compare a method that does this. The method was tested on data collected in the Central San Francisco Bay, where GPS surface-drifter deployments were conducted within the coverage of four 42 MHz radars over six days in October of 2008. Drifters were continuously deployed in these areas over the sampling days, providing 525 usable drifter measurements. Drifter and radar measurements were averaged into thirty-minute time bins. The three-day long-term averages from the sampling areas were then subtracted from the thirtyminute averages to remove biases associated with comparisons done with short, disjoint time-sample periods. These were then used to develop methods that give radar error or bias after the random fluctuations have been removed. Results for error estimates in this study are commensurate with others where random fluctuations have been filtered, suggesting they are valid. The estimated error for the radars in the SF Bay is low, ranging from −7.57 cm/s to 0.59 cm/s.