October 22, 2009 Meeting Summary
Vijay Tallapragada presented a status update and summary of HWRF H048 runs. Vijay explained that H048 represented the operational HWRF using phase 1 GFS upgrades, and it was run for a list of storms provided by NHC/Richard Pasch. For the 2008 hurricane season, Bertha (02L), Cristobal (03L), Dolly (04L), Fay (06L), Gustav (07L), Hanna (08L), Ike (09L), Omar (15L), Elida (06E), Fausto (07E), Genevieve (08E), Marie (14E), and Norbert (15E) were run. For the 2009 hurricane season, Bill (03L), Claudette (04L), Danny (05L), Erika (06L), Fred (07L), Henri (10L), Hilda (11E), Ignacio (12E), Jimena (13E), Linda (15E), and Olaf (18E) were run. While Felicia (08E) and Guillermo (10E) were requested by NHC for the 2009 hurricane season, GFS files were not available for these storms at the time. GFS reruns for these two storms are currently underway. Hilda and Ignacio were not requested by NHC, but were included.
Next Vijay presented track and intensity error as well as intensity bias plots for all storms from our testing run in the 2008 Atlantic, 2008 East Pacific, 2009 Atlantic and 2009 East Pacific. In the track and intensity error plots, operational HWRF was represented by red and H048 runs were represented by blue. For the bias bar charts, HWRF is represented by the blue bar and H048 by the maroon bar. The Atlantic 2008 runs had a total of 272 cases, while there were 128 cases in the 2008 East Pacific group, 73 cases for the 2009 Atlantic group, and 85 cases for the 2009 East Pacific group.
For the 2008 Atlantic storms, the track error for HWRF and H048 are very close with only a 5 nm degradation for H048 at 120h. Intensity error shows higher H048 error than HWRF at most forecast hours, but this degradation is only by a few knots, and mostly due to the higher intensity bias of H048. The intensity bias plot shows higher H048 values starting at 36h. However, the difference in HWRF and H048 biases is about 3kts at most. Vijay mentioned that the HWRF intensifies storms more with the new GFS which has more of an impact later on in the forecasts. For the 2008 East Pacific storms, H048 (the green lines in these plots) has higher track error values than HWRF up to 72h, after which, HWRF has higher errors than H048, especially at 120h. The intensity error plot shows HWRF with higher errors than H048 for all forecast hours except 120h. The intensity bias for 2008 East Pacific storms shows a negative HWRF bias throughout with H048 bias values less than 2kts for most of the forecast period.
For the 2009 Atlantic storms, H048 has a slightly higher track error than HWRF for most of the forecast period. The intensity errors for HWRF and H048 are very similar throughout. The intensity bias chart show H048 with slightly lower bias values than HWRF for the entire forecast period except at 48h. Bias values are also near or below 5kts until 96 and 120h. For the 2009 East Pacific storms, the HWRF and H048 had very similar values and exhibited the same overall pattern as the GFS. Here, H048 (the green line) is slightly worse than HWRF. The intensity error plot shows lower H048 error values until 48h when Vijay explained that HWRF's positive bias played a role with H048 intensity error values at 120h ~25kts more than those for HWRF. While H048 intensity bias values are slightly more than HWRF values, the H048 bias is less than 5 kts for every forecast hour except 120h. Here, Vijay mentioned that Jimena was the likely cause of such a large bias. Vijay concluded with a bar chart showing the frequency of superior performance for the 2008 Atlantic storms (on the left) and 2008 East Pacific storms (on the right). For these bar charts, the magnitude of superior performance was not considered, but simply which model was closer to observation. Overall, the HWRF and H048 showed very similar results for both the 2008 Atlantic and East Pacific.
Next, Janna O'Connor presented some individual storm statistics comparing H048 (in light blue) and HWRF (in red). The statistics shown were split into three categories: HWRF and H048 comparable, H048 performing better than HWRF, and H048 performing worse than HWRF. Most storms had comparable HWRF and H048 errors with only a few storms truly considered to have performed "worse" than HWRF. Janna noted that this list is not comprehensive and further investigation could be done by examining H048 statistics for individual storms on the restricted access HWRF website.
Bertha (02L) was the first storm shown in the comparable category. Except at 120h, HWRF and H048 track error were very similar, and intensity error values between the two differed by only a few kts. Fay (06L) and Gustav (07L) were also found to be comparable, while Hanna (08L) and Ike (09L) showed H048 track error values higher than HWRF values and larger errors as the forecast progressed. The same was true for Bill (03L) and Jimena (13E) with higher H048 track errors late in the forecast. Genevieve (08E) provided the best example for a case where H048 performed better than HWRF. H048 had lower track error by as much as 70 nm at 96h and a lower intensity error by as much as 12 kts at 24h. While Omar (15L) and Fausto (07E) were also shown in the H048 better category, the reduction in error from HWRF to H048 was not as large as that for Genevieve. Elida (06E) and Marie (14E) were shown as examples for H048 performing worse than HWRF. Marie's track error provides a better example than Elida's as H048 error values are higher than HWRF's by as much as 175 nm at 120h. Henri (10L) and Olaf (18E) are the final examples in the H048 worse than HWRF category. Both storms were short-lived and had track errors where HWRF and H048 were pretty similar. Henri's intensity error showed H048 with an 18 knot higher error value than HWRF at 36h. Olaf's intensity error difference between HWRF and H048 was not as large. Janna concluded by showing the HWRF restricted website and where the H048 graphics were located. Individual track graphics for Bill were then briefly discussed.