Evaluating temperature and humidity measurement biases in RS92 and RS41 radiosondes using radio occultation data

Abstract Upper-air temperature and humidity measurements from radiosondes are critical for monitoring atmospheric stability and climate trends. However, historical records can be affected by biases, inhomogeneities, and discontinuities, particularly during instrument transitions such as from RS92 to RS41. Ensuring long-term data homogeneity is therefore essential for reliable climate assessment. This study quantifies systematic biases in RS92 and RS41 radiosonde measurements at the Beltsville GRUAN Station, USA (2017–2020), and compares them with COSMIC-1 radio-occultation (RO) temperature and humidity profiles. Both daytime and nighttime profiles were analysed using mean differences, Bland–Altman plots, Taylor diagrams, and quadratic linear regression. Results indicate mean temperature biases between RS92 and RS41 of 0.08–0.39 K, with RMSE ranging from 0.47 to 0.93 K in the stratosphere (15–35 km). Bland–Altman analysis revealed temperature differences of −1–2 K and relative humidity biases of −9% to 5% across 1000–200 hPa. Comparisons with COSMIC-1 revealed strong correlations ( r = 0.98–0.99, p < 0.05) and normalised SDs of 0.15–0.6 K for temperature and relative specific humidity errors of 5%–25%, with larger deviations occurring in the upper troposphere where absolute humidity is low. These findings underscore the importance of global, long-term monitoring and calibration to enhance the quality of radiosonde data, which is essential for understanding atmospheric processes and improving climate change detection.