Evaluation of the GFDL 25km resolution Global Atmospheric Model for tropical cyclone prediction

This study utilizes a forecasting version of the high-resolution atmospheric model (HiRAM) developed at the Geophysical Fluid Dynamics Laboratory (GFDL). This version features the finite-volume core on a cubed-sphere grid, a modified shallow convection scheme, and 32 vertical model levels. It also includes a vortex breeding routine, a 4-D initialization large-scale nudging scheme, and a SST 'anomaly persistence' scheme. The model domain features 361 x 361 grid points in the horizontal, with an average grid spacing of approximately 25km. All simulations are performed at Oak Ridge National Laboratory on their high performance computing system (Jaguar). At the time of this writing, Jaguar is the most powerful supercomputer in the world for open scientific use, and features over 125,000 processing cores, 62 TB of memory, 600 TB of disk space, and a peak performance of 263 teraflop/s.

In the first part of this work, once-daily, 00Z simulations are performed between the period July 20, 2009 through September 31, 2009. These simulations are comprised of a large-scale nudging and storm-scale vortex-breeding step between t=-1 d and t=0, followed by a five-day model integration with these options turned off. All HiRAM tropical cyclone (TC) track and intensity forecasts are compared to a basket of operational limited-area and global dynamical models for both the Atlantic and Eastern Pacific ocean basins. Initial results indicate that TC track forecasts for the Atlantic basin are skillful relative to a majority of models, especially at longer lead times (four and five days). While there is little improvement in the Atlantic TC intensity forecast relative to other dynamical models, the significant intensity bias suggests that future HiRAM simulations at higher resolutions may improve intensity forecasts. HiRAM performance in East Pacific ocean basin is not as good when compared to TC forecast statistics from the Atlantic.

In the second part of this study, a unique ensemble-type approach is utilized to produce basin-wide TC forecasts on the intraseasonal timescale. Two, once-daily, 00Z HiRAM configurations are integrated forward in time for four weeks for the same start dates as specified in part one. Eight different basin-wide TC number and total storm day counts are constructed using a 21-day window starting at t=0 for each realization. For example, window one (eight) extends from t=0 (t=7 d) to t=21 d (t=28 d). This experimental design results in 2 x 8 = 16 ensemble members for a given 21 day window, provided that the window start day falls between July 28, 2009 and September 23, 2009. Ensemble averages of the 21-day windows of TC number and total storm days in both the Atlantic and East Pacific basins are compared to observed 2009 TC storm count and total storm days as well as 1978-2007 climatology. Since results from these simulations may be sensitive to the TC tracking algorithm, the formulation of the TC tracking algorithm is presented. It is argued that the TC tracker is robust, and is capable of differentiating between a warm core (tropical) vortex and a cold core (extratropical) vortex.