January 7, 2010 Meeting Summary
Vijay Tallapragada presented preliminary results from the FY2010 HWRF experiments. The experiments run included H050, which was the new baseline based on the current operational HWRF and included all bug fixes; H051, which used the H050 configuration and included initialization changes and additional data in the GSI based on Qingfu's work; H052, which also used the H050 configuration and included surface physics changes based on work done by Young Kwon; and H053, which also used the H050 configuration with gravity wave drag (GWD) turned on and run by Zhan Zhang. Storms run as test cases included: Cristobal, Dolly, Fay, Gustav, Hanna, Ike, and Omar from the 2008 Atlantic; Elida, Fausto, Genevieve, Marie, and Norbert from the 2008 East Pacific; Bill, Claudette, Danny Erika, Fred, Henri, and Ida from the 2009 Atlantic; and Felicia, Guillermo, Hilda, Ignacio, Jimena, Linda, Olaf, and Rick from the 2009 East Pacific. Of all these storms, there were seven storms designated "priority cases". They included Fay, Gustav, Hanna, Ike, Elida, Norbert, and Felicia. For H050, all storms were run and have been completed. For H051, H052, and H053 experiments, the priority cases were first run and based on preliminary results, the other storms were also run. Thus far, all storms for all experiments have been run with the exception of a few storms for H052. Vijay mentioned that the intended deadline for these runs was January 15th, and this testing is well ahead of schedule. Results presented by Vijay were only from the priority cases mentioned above.
First off, Vijay showed track and intensity errors from the 2008 Atlantic priority storms. In each plot, HWRF is the operational HWRF (in red), H48N is the current operational HWRF using the new GFS (in dark blue), H050 is the new baseline (in green), H051 is the experiment with new initialization (in yellow green), H052 is the experiment with new surface physics (in pink), and H053 is the experiment with GWD (in orange). For the track error plot for all four storms (Fay, Gustav, Hanna, and Ike), Vijay pointed out that H051 had a consistently better performance than the other experiments up to 96h where it was slightly degraded. Intensity error for all four storms shows increased error values for H050, H051, and H053 due to bug fixes, which also increased the positive intensity bias. After 48h, H052 had the lowest intensity error values.
Looking at individual storms, Vijay presented track and intensity errors for Fay first. Here, H051 had lower track errors, especially after 48h, due to the increase of GSI data in the landfalling region of the storm. Vijay mentioned that Fay was the worst performing storm with regards to intensity. H050, H052, and H053 all showed increase in intensity errors early on. Vijay mentioned that all the surface physics changes do not affect storms only over water and only take effect when winds are greater than 60 kts (Fay was only a tropical storm). The model also showed the storm over water when it was making landfall which also affected intensity. For Gustav, there was a slight degradation in track by H050, H051, H052, and H053 compared to H48N. After 48h, H051 had lower intensity errors than other experiments. H052 showed lower intensity errors for Gustav after 60h when compared to the other experiments. For Hanna's track error, H051 showed the lowest values from 24h onwards. The H050, H051, H052, and H053 experiments showed no significant improvement or degradation in Hanna's intensity error. H052, however, did show improved intensity over H050 beyond 48h. For Hurricane Ike, all configurations of H050 showed a degradation in track after 60h, however, Ike's intensity error was lowered significantly by H052.
Looking at the numerical values for average track errors for all four storms, Vijay noted that H051 had the lowest track error compared to the other H050 configurations from 24h onwards. For the relative errors compared to H48N, H051 showed a 10% improvement at 72h and all the H050-based tests were better than the H050 configuration itself. Looking at the frequency of superior performance, H051 did the best of all the H050 tests. For the numerical values for individual storms, H48N showed higher error values than HWRF indicating a degradation from using the new GFS. Also, H050 had higher error values than H48N indication a degradation due to the bug fixes.
Next Vijay looked at track and intensity errors for priority cases Elida and Norbert from the 2008 East Pacific. For both storms combined, H051 and H053 both showed an improvement in track. Vijay mentioned here that H053, which used GWD, showed neutral results in the Atlantic but improvements in the East Pacific. However, Vijay questioned whether GWD use was necessary because the error values for H051 were just as low as those for H053. Looking at intensity errors for both storms, H052 showed improved skill. For Elida, individually, the H050 experiments showed a slight degradation in track early on with H053 showing improved track later. Elida's intensity error showed H053 with the lowest values from 60h onward. For Norbert's track error, H053 did not improve skill past 48h. Instead H051 reduced the error values. H052 showed lower intensity error values for Norbert.
To conclude Vijay showed composite tracks for Ike on slide 36 of his presentation for the 2008 operational HWRF (upper left), H050 (upper right), and H48N (lower plot). He noted that the number of northward-moving tracks (in the black box) increased from the 2008 operational HWRF to H48N and was the most numerous for the 2010 baseline (H050). This contributes to the large errors seen for this storm. Vijay also noted that the simulation of the subtropical high was different for each of the three experiments shown, and HWRF tends to make the high too weak. Individual experiment tracks on the next few slides also indicate the more northerly track taken by H48N, H050, H051, H052, and H053 compared to HWRF. More findings on this will be presented in meetings to come.