Search the RAMMB website
(Sponsored by the Joint Hurricane Testbed)
An operational algorithm to estimate the confidence of the intensity forecasts from NHC’s primary intensity models and their consensus will be developed. The models include the statistical-dynamical Decay-SHIPS (DSHP) and Logistic Growth Equation Model (LGEM) and the early versions of the GFDL and HWRF coupled ocean-atmosphere models (GHMI and HWFI). The technique builds on the results of Bhatia and Nolan who demonstrated that the errors and biases of DSHP, LGEM, and GFDL have significant systematic variability as a function of a number of storm environmental variables that are available in real time, including the magnitude of the vertical shear, the direction of the shear, the initial intensity, and the maximum potential intensity. The intensity model error will be estimated from a linear combination of these predictors, supplemented with other variables. These include additional synoptic parameters, inner core structure from infrared imagery and the eye diameter and radius of maximum wind parameters from the Automated Tropical Cyclone Forecast (ATCF) system, ocean input from the sea surface temperature and oceanic heat content, the spread of the individual intensity models forecasts, and the recent performance of each model from times before the forecast time. Versions will be developed for the Atlantic and the combined East/Central Pacific. This algorithm will be referred to as the Prediction of Intensity Model Error (PRIME) model. The PRIME model will be run at the end of the SHIPS model script.
Assuming the model errors can be reliability estimated, the output from the PRIME model will be used to develop a corrected consensus forecast, which will be an unequally weighted combination of DSHP, LGEM, GHMI and HWFI forecasts. The error analysis will also be used to provide guidance for improvements to the DSHP and LGEM models.
An experimental version of PRIME will begin running at CIRA/CSU starting Aug 1, 2014.