12B.7 Ocean Model Impact Study for Operational Hurricane Blanca Forecasting

Thursday, 7 June 2018: 9:45 AM
Colorado B (Grand Hyatt Denver)
Hyun-Sook Kim, IMSG and NOAA/NWS/NCEP/EMC, College Park, MD; and A. Mehra and D. Iredell

For the 2015 Eastern Pacific hurricane season, ocean was a dominant source of thermal energy for TCs that rapidly intensified to major hurricanes over warm waters and favorable atmospheric conditions. In this broad context, accurate ocean coupling is essential for TC prediction models to capture the evolving intensity fluctuations. In this study, we investigate real-time forecast performance of Hurricane Blanca (2015) as a case study, using coupled Hurricane Weather Research Forecast (HWRF) systems. In the present configurations for the U. S. National Centers for Environmental Prediction (NCEP)’s HWRF model, two 3D ocean models – Princeton Ocean Model (POM) which is operational and HYbrid Coordinate Ocean Model (HYCOM) for the prototype pre-operational system are available for two-way dynamic coupling. Each ocean model solves the same free-surface, primitive governing equations, but with initial conditions (ICs) and boundary conditions (BCs) from climatology temperature and salinity for POM and data-assimilative analysis for HYCOM. Our study focuses on the impact of these ICs and BCs on air-sea interactions and on TC predictions for Hurricane Blanca.

Results show that the data assimilative based ICs represent more realistic ocean states that were of atypically warmer and deeper upper layers, than the climatology-based ICs. Independent observations from Argo floats and remotely-sensed products individually supported these up-to-date ICs. In spite of using the same parameterization for the atmospheric model component as in the operational version, HYCOM coupling demonstrated superior skills in intensity forecasts. Track forecasts with HYCOM coupling were mixed, improvements for early lead times, but degradation for late lead times. Intensity forecast improvements were, notably, persistent for all 33 cases, showing reduction of ≤13.5% for mean absolute error and of ≤75% for the bias. Better performance was especially demonstrated for the intensification cycles. We present and discuss these results in detail.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner