Kun Gao

My research focuses on improving the prediction of intensity, track and the heavy rainfall of hurricanes in high-resolution numerical models. I also conduct basic research to understand how the storm-scale processes (e.g., convection, large eddies) affect the hurricane structure and intensity changes, and also how the large-scale atmospheric variability affects the basinwide and regional-scale hurricane activity.

My recent research highlights are as below.

  1. I have been leading the development of the GFDL T-SHiELD experimental hurricane forecast system, which shows excellent hurricane track prediction skill (10-20% better accuracy than the current U.S. operational models GFS and HWRF). This is particularly evident in the 2022 hurricane season, in which the realtime T-SHiELD correctly predicted the landfall location of devastating Hurricane Ian 5 days before the actual landfall and showed much more consistent track prediction than the operational GFS and HWRF.
  2. I have recently implemented an initial vortex correction procedure in GFDL T-SHiELD that addressed the longstanding poor vortex initialization issue in the model. The procedure adjusts the initial vortex intensity and size based on observations and significantly reduces the negative storm intensity bias in T-SHiELD out to 72 hours.
  3. I conducted the first study on the impact of explicit model convection on the hurricane track prediction in large-domain storm-resolving (3 km resolution) models, and highlighted the overlooked role of fine-scale explicit convection in affecting the synoptic scale flow and consequently hurricane track forecasting skill. 

Prediction of the catastrophic Hurricane Ian (2022; third-costliest US weather disaster on record) by GFDL SHiELD and T-SHiELD, in comparison with operations models around the globe

Hurricane Ian

Image credit: Tim Marchok