Does Assimilating PlanetiQ and Spire GNSS RO Bending Angles Improve HAFS Atlantic Hurricane Forecasts? An Evaluation in Support of the ROMEX Experiment

The cloud-resolving Hurricane Analysis and Forecast System (HAFS) became NOAA’s operational regional tropical cyclone (TC) forecasting model in 2023. HAFS’s Gridpoint Statistical Interpolation (GSI) data assimilation (DA) algorithm assimilates a rich set of remotely sensed and in-situ observations, including Global Navigation Satellite System (GNSS) radio occultation (RO) bending angle data via a one-dimensional forward operator. This presentation summarizes results from data denial experiments evaluating the impacts of assimilating commercial GNSS RO observations on HAFS version A (HAFS-A) model analyses and forecasts, in alignment with the objectives of the International Radio Occultation Working Group (IROWG)’s Radio Occultation Modeling Experiment (ROMEX). Cycled HAFS forecasts of four 2022 Atlantic hurricanes – Earl, Fiona, Ian, and Julia – are run independently in two configurations: (i) Control, assimilating RO bending angles from COSMIC-2 and other government-operated missions routinely used by global weather forecasting centers; and (ii) Add-Spire-PlanetIQ, like Control but also assimilating RO data provided by the PlanetIQ (~ 3K daily profiles globally) and Spire Global (~ 17K daily profiles globally) companies. These bending angles from PlanetIQ and Spire are processed by the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT) and subsequently distributed to ROMEX participants by the Radio Occultation Meteorology Satellite Application Facility (ROM SAF).

Results based on about 100 HAFS forecasts from each experiment show that compared to those of COSMIC-2 observations, PlanetIQ and Spire observation errors diagnosed from observation-minus-background (O-B) and observation-minus-analysis (O-A) statistics are smaller in the extratropical lower troposphere. Compared to the Control forecasts, Add-Spire-PlanetIQ forecasts have statistically significant smaller track and intensity forecast errors for some lead times and reduced short-range lower troposphere temperature and water vapor random errors against airborne reconnaissance mission dropsondes. However, the quality control screens in HAFS-GSI reject over 50 percent of RO bending angles within and near the boundary layer, primarily due to these observations having either large O-B residuals or tangent points in a portion of the background where super-refractivity conditions are likely. Our results show that commercial Level 1 RO data provided by Spire and PlanetiQ, when processed to bending angles by operational forecasting centers, can meaningfully improve TC forecasts when assimilated into HAFS. However, given the sensitivity of TC intensification to lower-tropospheric water vapor, the most beneficial impacts from assimilating a larger volume of commercial RO data in HAFS may not be realized unless the retention rate of lower-tropospheric RO data can be improved.