Intercomparison of total column water vapor trends from ground-based radiometry and multi-GNSS solutions

Tropospheric water vapor is a key component of Earth’s climate system and plays a central role in atmospheric processes such as cloud formation, precipitation, and the transport of heat through evaporation and condensation. Its behavior is closely tied to atmospheric temperature via the Clausius-Clapeyron relation, which states that the amount of water vapor (in saturated air) increases exponentially with rising temperature. For real water vapor changes from multi-year to decadal time periods, several studies have revealed deviations from this theoretical scaling at regional spatial scales, highlighting the need for robust observational data to better understand these variations.

In this contribution, we estimate total-column tropospheric water vapor trends over a five-year period for a comparative performance evaluation, using multiple observational techniques, including ground-based radiometers operating in the microwave and thermal infrared bands, multi-Global Navigation Satellite System (GNSS) solutions, and reanalysis data. Each technique exhibits unique advantages and limitations, and comparing their outputs provides valuable insights into the consistency of total column water vapor retrievals and their potential for sensor fusion and synergistic retrievals.

We conducted an intercomparison of the total column water vapor trends, to assess biases, identify potential sensor drifts, and evaluate the overall accuracy of the individual trend estimates. Basis of this analysis are water vapor retrievals over 2021 to 2026 from measurements of co-located radiometers and a six-station GNSS station network, which are part of the WegenerNet Open-Air Laboratory for Climate Change Research in southeastern Austria. To obtain total column water vapor estimates from infrared radiometers, we simulate clear-sky brightness temperatures in the respective frequency bands from reanalysis data and use gradient-boosted regression trees with additional predictors to approximate the relation between total column water vapor and brightness temperature. A similar approach is used for the microwave radiometer. Our multi-GNSS water vapor estimates are based on precise-point-positioning solutions for each of the six stations.

We find that processing choices and hyperparameters play a crucial role for the estimated short-term trends for both the radiometer retrievals and the GNSS estimates. While we see an overall agreement between the observational techniques in trend direction, significant differences remain. We discuss the possible causes of the differences and related options for improvement learned from this intercomparison.