Impacts of Implementing a New GNSS-RO Quality Control and Error Model on Global Numerical Weather Prediction at NOAA

The assimilation of Global Navigation Satellite System (GNSS) radio occultation (RO) bending angle (BA) observations has been shown to produce improvements in a variety of numerical weather prediction (NWP) forecast metrics. One potential way to further advance the impacts of assimilating BA observations is through improved quality control (QC) and specification of observation errors in the variational data assimilation. This work investigates the impacts of implementing a dynamic error model for the specification of observation errors and updating the statistical QC in the NOAA operational analysis and forecast system. The current system uses a statistical error model that varies with latitude and RO mission. The new error model uses a combination of fractional local spectral width (LSW) in the troposphere (below 10 km), standard deviation of the RO bending angle observations from an exponential fit to the observations between 40 and 60 km above 30 km, and statistical error between 10 and 30 km. The dynamic error model is derived from the three-cornered hat method (Sjoberg et al. 2022) and allows RO observations with lower uncertainty to be given larger weights in the variational data assimilation than observations with higher uncertainty. The new statistical QC investigated in this study compares departures of BA observations from the background to a three-cornered hat statistical error value. The new QC and error model have no dependencies on satellite type, latitude, or other parameters, meaning that all RO observations are treated the same. The new QC and error model are being tested within the NOAA Gridpoint Statistical Interpolation (GSI) system for generating analyses and forecasts using NOAA's Global Forecast System (GFS) model. Initial results and evaluation will be presented at the conference.