January 28, 2010 Meeting Summary
Vijay Tallapragrada presented results from HWRF FY2010 experiments involving combined upgrades. For the statistics plots presented, HWRF represents the operational HWRF, H050 is the new baseline runs using bug fixes and GFS Phase-1 data, and H054 represents H050 plus changes to initialization including additional data in GSI near the storm (H051) plus surface physics changes involving Cd/Ch coefficients calculated based on observations (H052) plus gravity wave drag parameterization (H053). Vijay reminded everyone that the test cases run included: Cristobal, Dolly, Fay, Gustav, Hanna, Ike, and Omar from the 2008 Atlantic season; Elida, Fausto, Genevieve, Marie, and Norbert from the 2008 East Pacific season; Bill, Claudette, Danny, Erika, Fred, Henri, and Ida from the 2009 Atlantic season; and Felicia, Guillermo, Hilda, Ignacio, Jimena, Linda, Olaf, and Rick from the 2009 East Pacific season. This totaled in about 610 cases. Vijay also mentioned that in addition to all runs for H054 being completed, runs for priority cases (2008's Fay, Gustav, Hanna, and Ike) were completed for HYC1 and HYC2 experiments. HYC1 used the HWRF bug fix configuration with HYCOM instead of POM and HYC2 combined HYC1 with H054.
Next Vijay presented statistics from the 2008/2009 Atlantic cases, which was a total of 315 cases. The track error plot for the 2008/2009 Atlantic showed H051 (in purple) with comparable or lower error values than H050 (in blue) throughout. On the next plot, the track error values for H054 (in blue) are comparable to those for H051, and again H054 has lower errors than H050 (in red) throughout, so track skill was retained. For the plot of average track errors related to H050, H054 showed about a 10% improvement from 24h onward compared to H050. H054 was also improved over H050 for errors relative to HWRF. Vijay then showed the plot of intensity errors for the 2008/2009 Atlantic storms. Here, H052 (in pink) had the lowest error values from 36h onward. For the intensity error plot with H054 included, Vijay noted that H054 retained the skill shown in the previous plot with the exception of a slight degradation at 12h and at 120h. He mentioned that this lost skill was likely due to Gustav but more investigation into this was ongoing. In the intensity bias plot, H054 has reduced bias values compared to H050, and this is also true for standard deviation values, except at 12 and 24h where H050 has lower values. For the intensity errors relative to H050 plot, H054 showed a 25% improvement over H050 at 72 and 96h; however, H054 also showed a 10% loss of skill at 12h. This was also the case for the error relative to HWRF.
Next Vijay presented error statistics plots for the 2008/2009 East Pacific storms, which was a total of 298 cases. The average track error plot showed H051 (in purple) and H053 (in orange) with the lowest error values. The next plot, showing H054 results, confirmed that H054 retained the skill seen with H051 and H053 and it had the lowest error values compared to H050. There was a 5-10% improvement seen with H054 in the plot of track error relative to H050, and for track errors relative to HWRF, there was a 20% improvement at 96 and 120h for H054 compared to HWRF. In the plot of intensity error for the 2008/2009 East Pacific storms, Vijay noted that H052 (in pink) performed the best after 48h. Early on, H052 had some of the highest errors. For the intensity error plot featuring H054, H054 retained the skill seen in the previous plot, but again there were increased errors seen prior to 48h in the forecast period. For the plot showing intensity errors relative to HWRF, H054 showed a 10-12% improvement later in the forecast period compared to H050, and this results was also seen in the plot showing intensity errors relative to H050.
Vijay concluded by mentioning that the combination of individual upgrades produced a retention of track forecast skill from the initialization changes (H051) and a retention of intensity forecast skill from surface physics changes (H052). While there was a loss of skill early on, an improvement was consistently seen later in the forecast period. The current experiments being run for HWRF FY2010 testing include priority cases for H055 (which is H050 plus the GFS Phase-2 data), priority cases for H056 (which is H054 plus the GFS Phase-2 data), and Atlantic cases for HYC2. The final configuration for H210, which is the version of HWRF slated for implementation, will depend on results from H056 or HYC3 (which involves H054 plus HYCOM plus the use of GFS Phase-2 data).