A novel machine learning-based approach for improving global correction of AIRS-derived water vapor satellite product

Jiafei Xu, Zhizhao Liu

Research output: Journal article publicationJournal articleAcademic researchpeer-review


Precise precipitable water vapor (PWV) observations play a crucially important role in weather and climate research. The Atmospheric Infrared Sounder (AIRS) on-board the Aqua spacecraft is a hyperspectral instrument that offers operational PWV products using infrared (IR) channels. However, the observational accuracy of AIRS-sensed cloudy-sky PWV products is much poorer than that of PWV retrievals under clear sky conditions. We present a new machine learning-based calibration model to improve the observational performance of operational AIRS-derived IR PWV satellite products under all sky conditions, which considers using several dependence elements correlated with remotely sensed IR PWV observations. The first-guess PWV estimates, obtained from the ERA5 reanalysis, are also utilized in the newly developed calibration approach. The ground-based GNSS-sensed PWV observations, collected in 2017 across the world, are used as the expected PWV output for the training of the newly proposed calibration model. The newly calibrated PWV data records considerably outperform operational AIRS-sensed PWV products, compared with independent PWV estimates from radiosonde observations during 2017–2020. The root-mean-square error of operational PWV satellite products decreases 19.05 % from 2.31 mm to 1.87 mm under cloud fraction (CF) = 0 condition (clear sky), 30.71 % from 3.68 mm to 2.55 mm under CF = (0,1] condition (cloudy sky), and 30.14 % from 3.55 mm to 2.48 mm under CF = [0,1] condition (all sky). The calibration method exhibits much higher RMSE reductions than previous approaches that do not utilize the first-guess PWV estimates, illustrating the effectiveness of our calibration approach. This research provides insights into calibrating satellite-sensed PWV estimates based on machine learning by jointly using ground-based and reanalysis-based PWV observations. The newly proposed calibration method has significant potential to calibrate operational PWV products from other satellite-borne sensors, in addition to the AIRS instrument.

Original languageEnglish
Article number103787
JournalInternational Journal of Applied Earth Observation and Geoinformation
Publication statusPublished - Apr 2024


  • AIRS
  • Calibration
  • ERA5
  • GNSS
  • Machine learning
  • Water vapor

ASJC Scopus subject areas

  • Global and Planetary Change
  • Earth-Surface Processes
  • Computers in Earth Sciences
  • Management, Monitoring, Policy and Law


Dive into the research topics of 'A novel machine learning-based approach for improving global correction of AIRS-derived water vapor satellite product'. Together they form a unique fingerprint.

Cite this