Aircraft Reconnaissance Dropsonde and HDOB Data Impact on GFS Tropical Cyclone Forecasts

Tropical cyclones (TCs) are the most damaging natural hazards that regularly impact the coastal and inland populations across the world. Accurate predictions of TCs rely on the accuracy of model initial states over the oceans where the TCs originate. NOAA and US Air Force targeted aircraft reconnaissance observations, such as those from the dropsondes, help fill the data gaps from satellite and other in-situ observations especially in the North Atlantic and North Eastern Pacific basins, and have a long history of real-time assimilation for operational models at NCEP. The high-density flight-level reconnaissance observations (HDOBs) have also been added recently into the NCEP Global Forecast System (GFS) version 16 (GFSv16) for operational assimilation since March 22, 2021. In this study we carried out GFSv16 control and denial experiments with and without the use of dropsonde and HDOB data to examine their impact on TC forecasts, for a large sample of selected TCs during the 2020 to 2023 hurricane seasons. In the control experiment (Ctrl) the dropsonde and HDOB data are used, as in the GFS operations. In the denial experiment (Deny) the dropsonde and HDOB data are not assimilated in the GFSv16. Preliminary results from 2020-2022 hurricane seasons indicated that the dropsonde and HDOB data helped to improve TC track forecasts for strong storms. The dropsonde and HDOB data impact on the TC track for 2023 is not very clear. There are improved forecasts for some TCs (for example, improved track forecasts for Hurricane Lee) but degradation in other TCs (for example, track forecasts for TC Idalia at long lead times). Efforts have been made at NCEP to optimally use the dropsonde data in the GFS data assimilation system to improve TC track forecasts (e.g. Sippel et al. 2022). Additional testing is carried out for improved use of the dropsonde data in the GFSv16 data assimilation system, such as options for considering near-storm environment under certain TC intensity thresholds, and results from these experiments will be presented.