Simultaneous Multiscale 4DEnVar in Hurricane Analysis and Forecast System (HAFS)

The evolution of hurricanes is complex, involving interactions across multiple scales. The movement of hurricanes is influenced primarily by large-scale synoptic scale environment, while inner-core small-scale convection affects intensity evolution. In this study, we further developed the simultaneous multiscale data assimilation (MDA) method and implemented it with the 4DEnVar DA system in the regional FV3-based Hurricane Analysis and Forecast System (HAFS). Different from early studies, a Fast Almost-Guassian Filtering (FAGF) technique (Kovesi 2010) is developed and implemented to decompose the HAFS ensemble perturbations. Then a model space spatial scale dependent localization (SDL) is implemented. This method allows all resolved scales to be updated simultaneously while assimilating all observations at once. This study aims to answer if and how the simultaneous multiscale 4DEnVar method impacts hurricane predication relative to the single-scale method.

To address the question, we first implement the simultaneous multiscale 4DEnVar DA method within the HAFS system and tested it for Hurricane Laura (2020). Next, investigations were performed to compare the simultaneous multiscale approach with two single-scale localization (SSL) configurations. In-depth diagnostics suggest that (1) MDA using SDL produces better inner-core warm-core analyses that improves the intensity predictions during the intensification period than both SSL. (2) MDA using SDL produces better synoptic scale analyses that improves the track predictions during the early period than small-scale SSL. (3) MDA using SDL produces better TC structures in the forecast when compared to observations at almost all levels as compared both SSL. We will present our detailed findings and in-depth diagnostics at the AMS conference.