The Utility of a Two-Dimensional Forward Model for Bending Angle Observations in Regions with Strong Horizontal Gradients

Abstract Assimilation of Global Navigation Satellite System (GNSS) radio occultation (RO) observations into numerical weather prediction models in environments with strong horizontal gradients of refractivity introduces potential errors if one calculates the synthetic observations with a forward model (operator) that is only one dimensional. Innovations (observation minus background) from numerical experiments calculated using the RO Processing Package two-dimensional (ROPP2D) operator and background forecasts from the Global Forecast System (GFS) during Atmospheric River (AR) Reconnaissance 2022 are compared to those using the operationally employed ROPP1D and NCEP bending angle model (NBAM) operators. Throughout all regions examined, the lowest biases and standard deviations (SDs) of the innovations in the lower troposphere were produced by the NBAM operator, though differences in how superrefraction quality controls are performed compared to the ROPP operators complicate this comparison. Only slight reductions in bias and SD were found when using the ROPP2D operator compared to the ROPP1D operator over these broad regions. However, within intense ARs, where horizontal gradients are often extreme, the ROPP2D produces the smallest biases and SD of all operators, peaking at 2% less in terms of SD compared to the ROPP1D operator in the lower troposphere with a very small bias. While the use of a two-dimensional forward model for RO observations only results in slightly better correspondence to the observations over broad regions, this effect is much larger for weather phenomena with strong horizontal gradients and is likely to improve numerical forecasts of extreme precipitation events associated with ARs, an impactful weather phenomenon that is challenging to forecast. Significance Statement Despite the quantity of Global Navigation Satellite System (GNSS) radio occultation (RO) observations increasing rapidly as new GNSS-RO satellites are launched, these observations are only minimally used in the lower troposphere by numerical weather forecast models. Using a two-dimensional method to simulate these observations in models shows a greater than 2% improvement at 3 km in height in and around atmospheric rivers (ARs) where variations in the atmosphere are relatively large in the horizontal direction. This suggests that operational weather forecasting systems should employ this two-dimensional method to make better use of GNSS-RO observations in the lower troposphere and improve analyses and forecasts of impactful rain- and snow-producing weather phenomena like ARs.