January 21, 2010 Meeting Summary
Hyun-Sook Kim presented HYC1 test results. For the plots shown in her presentation, Hyun-Sook explained that HYC1 (in blue) represented the HWRF baseline (H050) with HYCOM instead of POM. H050 (in green) was the new HWRF baseline coupled to POM. H051 (in pink) represented H050 plus the new initialization, H052 (in aqua) represented H050 plus surface physics changes, and H053 (in black) represented H050 plus gravity wave drag. The last model used for comparison was HWRF, in red, which is the operational version of HWRF without the new GFS. The cases used for this work were 2008 Atlantic priority cases Fay (06L), Gustav (07L), Hanna (08L), and Ike (09L). The results from HYC1 are compared to those from H050 and maximum differences are noted on the plots.
Next, Hyun-Sook showed intensity error plots for all four storms combined. For the average intensity error, HYC1 performed worse compared to H050 by 1kt at 12h; however, at 72h, HYC1 had lower error values compared to H050 by 5kt. For intensity skill relative to SHIFOR, HYC1 had lower error values than H050 by 18% initially and then degraded at 12h as H050 had lower errors than HYC1 by 12%. The intensity error standard deviation showed HYC1 with 0.7kt higher error values than H050 at 24h and 1.8 kt lower error values than H050 at 96h. Hyun-Sook also noted that HYC1 was better than H050 and HWRF at forecast times greater than 72h. The average intensity bias plot showed HYC1 with lower error values than H050 throughout the forecast period, especially at 72h by 5.4 kt.
Hyun-Sook then presented track error plots for all four storms combined, starting with average track error. HYC1 showed lower average track error values than H050 at 126h by 24kt. For the rest of the forecast period, all models had fairly similar error values. Again, most models had similar error values in the skill relative to CLIPER plot, with the exception of HYC1 having a lower error value than H050 by 4% at 120h. The track error standard deviation plot showed HYC1 with lower error values than H050 by 9 nm in the 84-120h range. Next, Hyun-Sook presented a chart showing track bias. The forecast hours were on the x-axis while each model was along the y-axis with its own bias value. Hyun-Sook noted that at 96h, most models showed a north/northeast bias with the exception of GFDL and HWRF. She mentioned that this was mostly due to Fay, Hanna, and Ike, who all had northward biases individually.
Hyun-Sook summarized this work by noting that the average intensity forecast errors for HYC1 are reduced compared to the baseline, H050. This is especially true beyond 24h. Also, the intensity forecast for HYC1 proved to be skillful, by as much as 14%, especially after 24 hours. The average track errors and skills are also improved for HYC1 compared to the baseline, except at 12h. For her future work, Hyun-Sook mentioned conducting experiments for HYC2, which is H054 (H051+H052+H053) coupled to HYCOM, for the four priority cases. She also planned to tune parameters which should produce better results.
To conclude, Hyun-Sook presented a few slides on the SST feedback investigation. The plots shown for this pertain only to Ike (09L). The first plot showed the simulated minimum pressure compared to the best (on the left) for POM (in blue) and HYCOM (in red), both coupled to HWRF. Hyun-Sook noted that for both ocean models, HWRF overestimated the minimum pressure, especially for POM. This is also shown by the plot on the right of the wind-pressure relationship. Next, graphs for the 2008090212 cycle of Ike were presented showing the minimum pressure (top left), maximum velocity (top right), sea surface temperature (SST) (middle left), latent heat flux (middle right), wind stress (bottom left), and storm track (bottom right) with the best track in green. For all these plots, H050 is shown by the blue line and HYC1 by the red line. Hyun-Sook mentioned that the plot showing minimum pressure gives a 1000-940 mb range for HYC1 and a 1000-890 mb range for H050, further illustrating HWRF's overestimation of pressure. It was also noted that the tracks for H050 and HYC1 are fairly similar, but values for SST are different. Hyun-Sook explained that it appeared that HWRF with POM responded to SST in a more linear fashion than HWRF with HYCOM. Similar observations were made for the 2008090618 cycle of Ike, where HYC1 had 1 degree cooler SST values than H050 initially. It was suggested that a map be presented showing the SST difference, and more work on this subject is underway.
Hyun-Sook also mentioned that statistics for individual storms could be found in her presentation, even though they were not explicitly discussed in her presentation.