This work investigates whether machine learning (ML) can offer an alternative to existing methods to map radio occultation (RO) products, allowing the extraction of information not visible in direct observations. ML can further improve the results of Bayesian interpolation, a state-of-the-art method to map RO observations. The results display improvements in horizontal and temporal domains, at heights ranging from the planetary boundary layer up to the lower stratosphere, and for all seasons.

Crowdsourced smartphone GNSS data were processed with a dedicated data processing pipeline and could produce millimeter-level accurate estimates of zenith total delay (ZTD) – a critical atmospheric variable. This breakthrough not only demonstrates the feasibility of using ubiquitous devices for high-precision atmospheric monitoring but also underscores the potential for a global, cost-effective tropospheric monitoring network.

In this study, a comprehensive multi-disciplinary dataset for tropospheric water vapor was developed. Geodetic, photogrammetric, and atmospheric modeling and data fusion techniques were used to obtain maps of water vapor in a high spatial and temporal resolution. It could be shown that regional weather simulations for different seasons benefit from assimilating these maps and that the combination of the different observation techniques led to positive synergies.

This study develops an innovative approach for the detection and prediction of foehn winds. The approach uses products generated from GNSS (Global Navigation Satellite Systems) in combination with machine learning-based classification algorithms to detect and predict foehn winds at Altdorf, Switzerland. Results are encouraging and comparable to similar studies using meteorological data, which might qualify the method as an additional tool for short-term foehn forecasting in the future.