Despite their importance in climate, NWP, and satellite retrieval applications, our knowledge of the respective error characteristics of various instrument types comprising the WMO global radiosonde network is limited, due largely to the lack of a “global” transfer standard for comparisons. In the past, attempts have been made to inter-compare different radiosonde types through special studies conducted at specific locations where a limited number of radiosondes were compared under specific atmospheric conditions. Studies also are available for which the global sonde performance was characterized using polar satellites as a transfer standard, but the relatively low horizontal and vertical resolution of the satellite data and lack of synchronization between the sun-synchronous satellite and synoptic radiosonde observations posed serious limitations. The relatively high vertical resolution, accuracy, and comparatively random temporal distribution of the global COSMIC RO observations offer a unique opportunity to characterize performance among the different radiosonde types on a global scale using COSMIC as a transfer standard.

In this study, the satellite-based global positioning system (GPS) radio occultation (RO) soundings and measurements from the Constellation Observing System for Meteorology, Ionosphere, and Climate (COSMIC) mission are used as a “transfer standard” platform to inter-compare the different sonde types. Launched in April 2006, COSMIC currently yields ~2000 all-weather profiles per day distributed uniformly around the globe. Vertical profiles of COSMIC refractivity soundings are inverted from GPS RO bending angles, and COSMIC temperature (T) and water vapor (W) soundings are retrieved using the refractivity profiles in conjunction with a one-dimensional variational scheme (UCAR) that uses NWP analyses as its first-guess. Retrieval biases from 8 to 25 km are typically better than 0.1 K as an ensemble, and for individual profiles within 0.5 K. In the lower troposphere the bias depends on latitude with the largest values in the Tropics. Overall, the bias in COSMIC sounding data varies smoothly in space, making comparison of different radiosonde instrument types meaningful in a relative sense.

Spatially and temporally collocated radiosonde and COSMIC observations from the past six months are used for the analysis. Collocated observations are provided through the NOAA Products Validation System (NPROVS), routinely operated by the Office of SaTellite Applications and Research (STAR) at NESDIS. The basic function of NPROVS is to apply a consistent protocol for monitoring and validating multiple operational and research satellite observations and derived weather products in preparation for next generation NPOESS data processing systems. NPROVS has been running routinely since 2007 and has been useful for characterizing “relative” performance among Advanced TIROS Operational Vertical Sounder (ATOVS), Microwave Integrated Retrieval System (MIRS), Infrared Atmospheric Soundings Interferometer (IASI), Geostationary Operational Environmental Satellite (GOES), Atmospheric InfraRed Sounder(AIRS), and COSMIC satellite derived products.

The following report summarizes radiosonde instrument performance using the collocated radiosonde and COSMIC observations within NPROVS. Differences between radiosondes include uncertainties with respect to sensor radiation corrections and the type of sensors used for humidity measurement. These differences and other pertinent issues concerning the characteristics of the radiosonde observations used in this study are reported. Statistical techniques include the Kolmogorov-Smirnov (K-S) tests to determine the significance of respective measurement differences among the major instrument types and groupings as identified in the study. The goal is that this work can lead to standardization among the different radiosonde types for improved impact in satellite weather and climate applications.