Advancing Methodologies for Uninterrupted, Basin-Wide Data Assimilation in the Hurricane Analysis and Forecast System (HAFS)

The Hurricane Analysis and Forecast System (HAFS), NOAA's next-generation FV3-based tropical cyclone (TC) forecasting system, features flexible model configurations for both operational and research purposes. In particular, HAFS can operate with a large, static analysis domain that permits the uninterrupted assimilation of measurements basin-wide. This feature opens new research directions for TC prediction. The current presentation discusses findings from a series of numerical experiments performed over the peak of the 2022 TC season. These experiments seek to optimize the use of satellite radiance measurements by performing DA over an extensive domain, which permits the continuous collection and self-cycling of innovation statistics needed to perform variational bias correction. This design choice is motivated by Knisely and Poterjoy (2023), who show that such schemes can be effective for HAFS domain configurations that extend over entire hurricane basins. Furthermore, DA requires careful estimation of background uncertainty, but current operational regional NWP systems often use relatively small ensemble sizes which can lead to sampling deficiencies. All HAFS experiments performed for this research use a combined 40-member HAFS ensemble and 80 additional ensemble perturbations from the GDAS system, which consequently reduces sampling errors. Finally, we implement a Bayesian DA technique called the local particle filter (LPF) for direct comparisons with an EnKF. Since the LPF does not rely on covariances alone when describing background uncertainty, this method presents advantages for DA applications that are strongly impacted by nonlinear model dynamics and measurement operators. Lastly, we examine strategies for combining ensemble analyses and forecasts generated from basin-wide DA experiments with forecast domain configurations that closely resemble operational HAFS model configurations.