NOAA Privacy Policy | NWS Disclaimer
N.O.A.A. logo HWRF banner image National Weather Service logo

September 24, 2009 Meeting Summary

Mark DeMaria presented his and Brian McNoldys work on preliminary HWRF diagnostics for tropical storm Erika 2009, which is part of the HFIP project. First off, Mark presented HWRF intensity and track forecasts for Erika 06L. For intensity, HWRF intensified the storm over and over for each forecast. In reality, Erika was weak and in an environment of high shear. For track, the HWRF took the storm on a more northwestward track and later recurved it. As Mark mentioned, the goal of the diagnostic work on Erika was to figure out why the model behaved in such a way.

Next, Mark presented the average error (in the plot on the left) and bias (on right) for Erikas vertical shear (850-200mb). For these plots, shear from the GFS analyses along the observed storm track was used as “ground truth”. For the average error, HWRF began with approximately 5 kt errors, which increased around 24 hours before dropping and increasing again to about 8kts towards the end of the forecast period. For the bias, the overall trend was an increase in the mean difference (in kts) as the forecast hour increased, which indicated too low vertical shear in the model environment. This could have been a contributing factor to HWRFs overintensification of Erika. Looking at the average error (on the left) and bias (on the right) for SSTs, Mark showed that while the error for HWRF around 48 hours in the forecast is not large, this value steadily increased throughout the forecast. The bias also showed low values early on, but an increasingly negative bias as the forecast went on. Mark mentioned that this could be a consequence of the track error, and thus, the overintensification by HWRF was most likely an atmospheric and not an ocean problem. Then, a plot of the forecast (on the left) and observed (on the right) moisture structure was shown. The 48h HWRF moisture pattern was overall pretty organized, while the observed moisture (obtained from NESDIS microwave satellite imagery) was more disorganized. Mark explained that the model more than likely initialized with much more organization than what was observed and this error propagated through the entire forecast.

Mark concluded by presenting the next steps in this diagnostic work. First off, a comparison of model and satellite precipitation estimates would be performed for early in the forecast period. Second, a comparison of the time evolution of the model and observed infrared cloud top temperatures would be conducted. Lastly, Mark and Brian would apply steering layer analysis to compare the model and observed vertical storm structure to help determine why the HWRF intensified Erika so much. This could also be carried out for other storms from this hurricane season.

Young Kwon then showed two slides from his current work with Cd and Ch values and intensity bias. The first plot showed a bar graph of intensity bias (in kts) for Bertha, Hanna, Gustav and Ike. Of note, the operational HWRF (HOPR) is in red and H5_5 (in yellow) is an observation-based value for Cd/Ch. Young noted that using H5_5 values for Cd/Ch decreased the intensity bias for these storms, especially later on in the forecast. Next, Young presented an intensity bias bar graph for Erika only, which used all 48h forecasts for the lifetime of the storm. In this plot, the H5_5 Cd/Ch value did not show as large a reduction in intensity bias. Instead of a reduction of around 20kts seen for Bertha, Hanna, Gustav and Ike, a reduction of only 2-3kts could be seen for Erika. Additional work on Cd/Ch values is currently underway.

Please e-mail comments, questions, or suggestions about the contents of this webpage to Janna O'Connor, at

Home | HWRF Main Page
EMC | NCEP | National Weather Service | NOAA | Department of Commerce