GEWEX Water Vapor Assessment: Validation of AIRS Tropospheric Humidity Profiles With Characterized Radiosonde Soundings.

The tropospheric water vapor profile record from the Atmospheric Infrared Sounder (AIRS) now spans over a decade, making it a valuable resource for climate studies. To be considered as a Climate Data Record, it is key that the ultimate performance of these observations are understood. The Global Energy and Water cycle Exchanges (GEWEX) water vapor assessment has been tasked with characterizing the current state of the art in water vapor products currently available for climate analysis. Within the scope of this exercise, water vapor profiles from AIRS have been assessed using collocated characterized in situ measurements of tropospheric water vapor between 2007 and 2012. We first show how previously published methods for correcting radiosondes can be applied to global records, which show high correlations to Global Climate Observing System Reference Upper‐Air Network (GRUAN) performance at pressures between the surface and 250 hPa. We go further and show the first comparison of uncertainties from both the newly created Characterized Radiosonde Measurement (CRM) and GRUAN data sets. Global estimates of AIRS water vapor profile (wet/dry) biases relative to GRUAN and CRM are within 6 ± 0.3% ppmv and 15 ± 0.1% ppmv below 300 hPa, respectively. The CRM record allows latitudinal analysis for the first time, which when examined shows sensitivity to changes in absolute concentration due to large‐scale circulation in the International Tropical Convergence Zone. This paper advances the use of state‐of‐the‐art in situ records for characterizing absolute performance, recognizing that long‐term stability needs further research. Key Points: Altitudes below 250‐hPa corrected operational radiosondes show similar uncertainty performance to GRUANAIRS tropospheric water vapor biases relative to GRUAN are within 6 ± 0.3% ppmv below 300 hPaEstimated collocation uncertainty in the Northern Hemisphere shown to reduce to below 1% ppmv for calculated yearly biases [ABSTRACT FROM AUTHOR]