Observing atmospheric rivers using multi-GNSS airborne radio occultation: system description and data evaluation

October 1, 2025

Airborne radio occultation (ARO) allows research aircraft to act as storm-targeted vertical profilers of atmospheric rivers, providing more than three times the number of atmospheric profiles. A recent paper, “Observing atmospheric rivers using multi-GNSS airborne radio occultation: system description and data evaluation”, published in Atmospheric Measurement Techniques, was led by Bing Cao (IGPP/SIO) and Jennifer Haase (SIO/CW3E), with contributions from Michael Murphy (SIO/NASA Goddard/UMBC) and Anna Wilson (CW3E). The study presents a four-year ARO dataset (2018–2021) from Atmospheric River Reconnaissance (AR Recon) missions, spanning the central Pacific, where atmospheric rivers gather tropical moisture, and the eastern Pacific, where they make landfall and affect the U.S. West Coast. This work supports the Atmospheric Rivers and Extreme Precipitation Research, Prediction, and Applications, and Novel Observations priorities in CW3E’s 2025-2029 Strategic Plan by advancing our capability to observe critical parameters in the atmosphere and use these observations in operational forecasts. The results show that ARO has become a mature and reliable technology, providing dense, storm-targeted observations that strengthen our ability to understand and forecast atmospheric rivers

In total, 1,734 multi-GNSS occultation profiles were collected over ~266 flight hours, typically 5–7 profiles per hour or 35-50 profiles per flight, with ~80% of profiles inside the ±3 h data-assimilation window around 00:00 UTC. The ARO retrieval combines GPS/GNSS precise positioning, carrier-phase residuals, and a tailored partial-bending/Abel inversion to produce profiles of bending angle, refractivity, and dry-air pressure/temperature suitable for research and NWP. Sampling resolves ~400 m vertically between 5–10 km altitude range. The horizontal along-path integration is typically ~200–300 km centered on the tangent points. Because tangent points drift, profiles extend sampling well beyond the flight track—typically ~400–600 km, and up to ~700 km in deep cases—capturing cross-AR structure that complements dropsondes. Figure 1 illustrates a representative data distribution of one AR Recon flight: the clockwise flight track, dropsonde release points, and the projected tangent-point paths for setting (solid) and rising (dashed) occultations are overlaid on IVT magnitude and vectors; symbols denote the lowest-point altitude (<3 km, 3–6 km, >6 km), short line segments highlight the ray-path portion contributing 50% of excess phase.

A detailed evaluation of ARO data confirms its accuracy for use in numerical weather prediction models. When compared with ERA5 reanalysis, refractivity differences are small—biases are less than 0.5% with standard deviations under 1.5% above 4 km (reaching a minimum of about 1% near 8 km), increasing to around 2.8% below 4 km. Relative to dropsondes and ERA5, the data show slightly positive biases at higher altitudes and near –2% close to the surface, consistent with known super-refraction effects. These statistics, summarized in Figure 2, include results for all profiles as well as broken down by GPS, GLONASS, and Galileo constellations. To maximize the value of ARO, a specialized assimilation operator has been developed that uses the more primitive refractive bending measurements while accounting for its unique asymmetric geometry (Hordyniec et al. 2025). Multiple assimilation experiments using this operator have demonstrated positive impacts on atmospheric analyses and short-range forecasts (Do et al. 2025; Haase et al. 2021). The full dataset, including maps and CDAAC-style atmPrf files, is openly available through the UCSD Library and the project’s website.

Figure 1. (Figure 5 from Cao et al. 2025). Overview of IOP04 (4 Feb 2020): flight track (thin black line), dropsonde releases (black circles), and setting/rising ARO tangent-point paths (solid/dashed blue lines) over IVT, with symbols marking lowest-altitude class and short segments indicating the ray-path portion contributing 50% of excess phase.

Figure 2. (Figure 12 from Cao et al. 2025). Summary of ARO–ERA5 refractivity differences—mean and standard deviation versus height (all profiles and by constellation)—showing <0.5% bias and <1.5% SD above 4 km and broadly consistent performance across GPS, GLONASS, and Galileo.

Cao, B., Haase, J. S., Murphy Jr., M. J., & Wilson, A. M. (2025). Observing atmospheric rivers using multi-GNSS airborne radio occultation: system description and data evaluation. Atmospheric Measurement Techniques, 18(14), 3361–3392. https://doi.org/10.5194/amt-18-3361-2025

Hordyniec, P., Haase, J. S., Murphy, M. J., Jr., Cao, B., Wilson, A. M., & Banos, I. H. (2025). Forward modeling of bending angles with a two-dimensional operator for GNSS airborne radio occultations in atmospheric rivers. Journal of Advances in Modeling Earth Systems, 17(4), e2024MS004324. https://doi.org/10.1029/2024MS004324

Do, P.-N., Haase, J. S., Baños, I. H., Hordyniec, P., & Cao, B. (2025). Impact of airborne radio occultation observations on short term precipitation forecasts of an atmospheric river. Geophysical Research Letters, 52(13), e2025GL115639. https://doi.org/10.1029/2025GL115639

Haase, J. S., Murphy, M. J., Cao, B., Ralph, F. M., Zheng, M., & Delle Monache, L. (2021). Multi-GNSS airborne radio occultation observations as a complement to dropsondes in atmospheric river reconnaissance. Journal of Geophysical Research: Atmospheres, 126(21), e2021JD034865. https://doi.org/10.1029/2021JD034865