Quality assessment of COSMIC/FORMOSAT-3 GPS radio occultation data derived from single- and double-difference atmospheric excess phase processing

This study evaluates the quality of GPS radio occultation (RO) atmospheric excess phase data derived with single- and double-difference processing algorithms. A spectral analysis of 1 s GPS clock estimates indicates that a sampling interval of 1 s is necessary to adequately remove the GPS clock error with single-difference processing. One week (May 2-8, 2009) of COSMIC/FORMOSAT-3 data are analyzed in a post-processed mode with four different processing strategies: (1) double-differencing with 1 s GPS ground data, (2) single-differencing with 30 s GPS clock estimates (standard COSMIC Data Analysis and Archival Center product), (3) single-differencing with 5 s GPS clocks, and (4) single-differencing with 1 s GPS clocks. Analyses of a common set of 5,596 RO profiles show that the neutral atmospheric bending angles and refractivities derived from single-difference processing with 1 s GPS clocks are the highest quality. The random noise of neutral atmospheric bending angles between 60 and 80 km heights is about 1.50e-6 rad for the single-difference cases and 1.74e-6 rad for double-differencing. An analysis of pairs of collocated soundings also shows that bending angles derived from single-differencing with 1 s GPS clocks are more consistent than with the other processing strategies. Additionally, the standard deviation of the differences between RO and high-resolution European Center for Medium range Weather Forecasting (ECMWF) refractivity profiles at 30 km height is 0.60% for single-differencing with 1 and 5 s GPS clocks, 0.68% for single-differencing with 30 s clocks, and 0.66% for double-differencing. A GPS clock-sampling interval of 1 s or less is required for single- and zero-difference processing to achieve the highest quality excess atmospheric phase data for RO applications.