88th Annual Meeting (20-24 January 2008)

Monday, 21 January 2008
Improving Hurricane Intensity Forecasting via Microphysical Parameterization Methods in a Mesoscale Model
Exhibit Hall B (Ernest N. Morial Convention Center)
Cerese Marie Albers, The Florida State University, Tallahassee, FL; and D. T. N. Krishnamurti
The goal of this study has been to improve mesoscale modeling of hurricane intensity. This is done via the comparison of field campaign observations of Hurricane Erin 2001 from The Fourth Convection and Moisture Experiment (CAMEX-4) and Hurricane Dennis 2005 from the Tropical Cloud Systems and Processes (TCSP) mission with simulated results of improved microphysical parameterization in a mesoscale model utilizing the Krishnamurti, et al (1991) technique of Rain Rate Initialization.

            First a 72-hour control run of Hurricane Erin 2001 in the Weather Research & Forecasting (WRF) model was made without altering any dynamics or physics packages to see how the model performed with respect to the storm. The seventy-two hour model run was made beginning at 00Z on September 8, 2001 and terminating at 00Z on September 11, 2001, adequately capturing the developing and mature stages of the hurricane, as well as the time period during which it was sampled by the ER-2 and DC-8 aircraft during CAMEX-4 on September 10, 2001. Next, a run of Hurricane Erin 2001 in the WRF model was made utilizing Rain Rate Initialization during the first 24 hours and a 48 hour forecast was made after that. These results were compared with the National Hurricane Center's (NHC) Best Track and Intensity Analysis of Hurricane Erin 2001. Seeing marked improvement with Rain Rate Initialization alone, the microphysics aspect was then combined with Rain Rate Initialization to provide optimal storm intensity forecasts that can be compared with NHC and CAMEX-4 observational data.

A series of 25 microphysical experiments were designed to isolate the individual effects of altering one microphysical parameter at a time on the hurricane's intensity (see Table 1). Building off of the results from the study of Pattnaik and Krishnamurti (2006), specific microphysical parameters were selected and would be altered based on their effectiveness in prior microphysical experiments (detailed in a literature review). This was done by changing the melting of snow, the melting of graupel, the evaporation of rain water, the fall speed of snow, the fall speed of graupel, and the intercept parameter for graupel, one parameter at a time. Each parameter was taken at 100%, 75%, 50%, 25% and 0% of its maximum value in the model, one experimental value at a time, and one parameter at a time, to form 25 experiments.  Each of these experiments began after the initial 24-hour period where Rain Rate Initialization (RRI) occurred (or beginning at 09-09-2001 at 00Z) and terminates after 48 hours (at 09-11-2001 at 00Z). This 48-hour time period was chosen for the following reasons: First, running the 25 experiments for more than the full 48 hours after RRI would be highly computationally intensive and would take up a lot of computational resources and time. Second, this captures enough of the mature stage of the hurricane that the experiments performed occur during a relevant time. Third, a full 72-hour forecast (post- RRI) was made once an optimal combination of the five parameters was deduced using an RMS Error optimization technique.

This technique involved taking all of the 25 experiments and finding which percentages of each variable had the greatest effect on improving the intensity estimate of Hurricane Erin 2001 through RMS Error. The RMS Errors were computed from comparing the model output of each experiment to observed data. Observations included wind speed, minimum sea level pressure, TRMM (or CMORPH) rainfall data for rain band structure and rainfall estimation, track, temperature anomaly in the warm core (if possible), and vertical profiles of the hydrometeor distribution (as compared with the model control run). These RMS errors show, through a unique calculation, what the optimal combination of microphysical parameterization should be. Then a final model run was made utilizing this optimal combination to produce a Rain Rate Initialized, microphysically parameterized, 72-hour forecast of Hurricane Erin 2001 which was compared at the appropriate corresponding times with observations made during CAMEX-4 (as shown below in Fig 2). Observations from the field campaign included a combination of aircraft and satellite observations including hydrometeor profiles, brightness temperatures from AMPR (Advanced Microwave Precipitation Radiometer), radar reflectivity, and dropsonde measurements. The result was also compared with the model control run and the pure RRI control run without microphysical parameterization to show the improvement in the intensity forecast made by combining the best options.

The implications of this study will be three-fold. First, the value of using RRI in mesoscale models is proven. Second, the value of specific, useful, microphysical parameterization in hurricanes is shown to make a marked improvement in hurricane intensity forecasting. Third, the use of valuable field experiment data for validation of experimental modeling techniques is shown.

In addition to the Hurricane Erin (2001) experimentation the same was done for Hurricane Dennis (2005). Further results are shown from comparisons of the two storms showing which microphysical parameterizations worked optimally in both storms and strengthening the evidence that RRI and proper microphysical initialization in mesoscale hurricane modeling are both useful and effective techniques for improving hurricane intensity forecasting.

Supplementary URL: