26th Conference on Hurricanes and Tropical Meteorology

Wednesday, 5 May 2004
The Advanced Objective Dvorak Technique (AODT): Latest Upgrades and Future Directions
Richelieu Room (Deauville Beach Resort)
Timothy Olander, CIMSS/Univ. of Wisconsin, Madison, WI; and C. S. Velden and J. P. Kossin
Poster PDF (95.3 kB)
The advancement of an objective scheme for determining tropical cyclone (TC) intensity estimates using geostationary, infrared satellite data has been in progress at the University of Wisconsin-Madison/Cooperative Institute for Meteorological Satellite Studies (UW-CIMSS) since the mid 1990s. This algorithm has progressed from a relatively simple, manually-interactive routine limited to analyses of storms of hurricane/typhoon strengths only, to a completely automated algorithm capable of providing intensity estimates for all TC lifecycle stages. Future plans for the AODT include a careful error analysis to develop situation-dependent confidence indices that can be provided along with the intensity estimates. Eventually, the AODT will be merged with other existing satellite-based TC intensity estimation algorithms, using multiple satellites instruments and spectral channels, into an integrated "expert system" that will exploit the strengths of each technique.

The release of the latest version of the AODT includes several major performance improvements and methodology additions: 1) the ability to operate the AODT in a completely automated manner has been implemented. This technique relies on OC short- term position forecasts as a first guess, with a Laplacian analysis of the cloud top temperature field used to reposition the storm center, if necessary, 2) modification of the scene identification scheme has significantly improved scene classification during formation and weakening stages, especially with the eye versus curved band classifications, 3) additional rules/constraints on the estimates have been implemented. These rules primarily preserve the consistency of the estimates and allow the AODT to more closely resemble the subjective method TC forecasters are accustomed to, 4) the experimental Tropical cyclone Intensity Estimate (TIE) model, a regression-based intensity estimation algorithm, has been added to the AODT package to provide TC forecasters with another analysis tool, 5) a regression-based, latitude-dependent bias adjustment has been added to correct the AODT intensity estimates of MSLP based upon the latitudinal storm position.

Comparisons of AODT intensity estimates with aircraft reconnaissance MSLP measurements have shown the AODT algorithm to possess accuracy on par with those estimates obtained by TC Operational Centers (OC) using the subjective Dvorak Technique (DT). RMS errors of 10.04mb for the Automated-AODT versus 10.65mb for the Operational Center DT estimates were achieved for a dependent, homogeneous sample of 29 storms (1630 matches) between 1995 and 2002. Manually positioning the storm center location for the AODT analyses lowered the RMS error by 0.5mb.

Supplementary URL: