Evaluation of Retrospective Hafs Forecasts with a Large Airborne Radar Dataset

The Hurricane Analysis and Forecast System (HAFS) was implemented operationally during the 2023 North Atlantic hurricane season. To prepare for operational implementation, three-year retrospective forecasts were run for the North Atlantic and eastern North Pacific basins to assess the forecast skill of HAFS. In this study, we make use of airborne radar observations to explore tropical cyclone (TC) structure in these retrospective forecasts, advancing evaluation beyond the “standard” metrics of track and intensity skill. For this comprehensive evaluation, we use the Tropical Cyclone Radar Archive of Doppler Analyses With Recentering (TC-RADAR) dataset, which contains post-processed three-dimensional analyses of TC structure from NOAA P-3 Tail Doppler radar data from flights from 1997 to 2022. For this study, we utilize the data collected by P-3 reconnaissance missions into North Atlantic TCs from the extremely active 2020-2022 hurricane seasons to form a homogeneous dataset for comparison with both operational versions of HAFS: HAFS-A and HAFS-B. We compare HAFS-A and HAFS-B with TC-RADAR through several methods, including comparison of composite wind fields, calculation of structure metrics for the full homogeneous dataset, and comparisons of the distributions of vertical velocity and other fields. This analysis provides three-dimensional insight into both the overall biases of HAFS compared with observations, as well as the differences between the two operational versions. We also highlight observational comparisons for a smaller set of model physics experiments, showing how different planetary boundary layer and microphysics parameterizations affect the modeled TC structure in HAFS. This model-radar comparison will form a framework for future evaluation of both operational and experimental versions of HAFS and other hurricane models and will also help inform upgrades to model physics for improved forecasts of TC structure.