A statistical-dynamical ensemble was developed to provide tropical cyclone (TC) intensity guidance, called Statistical Prediction of Intensity from a Consensus Ensemble (SPICE). SPICE utilized a combination of two statistical-dynamical models, the Logistic Growth Equation Model (LGEM) and the inland decay version of the Statistical Hurricane Intensity Prediction Scheme (SHIPS). Both of these models were run with environmental parameters provided from three different dynamical models, producing a total of six forecasts of TC intensity. The three dynamical models chosen to provide input to the statistical-dynamical models were the Global Forecast System (GFS), the Hurricane-Weather Research and Forecasting (HWRF) model, and the Geophysical Fluid Dynamics Laboratory (GFDL) hurricane model. SPICE averaged the three forecasts of SHIPS to form an unweighted SHIPS consensus, and averaged the three forecasts of LGEM to form an unweighted LGEM consensus. SPICE then combined the two unweighted consensus forecasts into one weighted consensus, with the weights determined empirically from the 2008-2010 SHIPS and LGEM samples, to produce a single forecast on tropical cyclone intensity. These weights favored the SHIPS consensus in the early time periods, shifting to the LGEM consensus being weighted more heavily after about 36 hours. SPICE has been run retrospectively since 2008 and in real-time demonstrations since 2011 as part of the Hurricane Forecast Improvement Project (HFIP). During the real-time demonstrations, SPICE produced lower mean absolute errors than either parent statistical-dynamical model, SHIPS or LGEM, at longer forecast times. SPICE also produced lower mean absolute errors than the parent dynamical models, GFS, HWRF, and GFDL, at longer forecast times.
The discontinuation of GFDL as an operational model in July 2017 necessitated a re-design of the SPICE ensemble. Experiments with other global and hurricane models, including the Hurricanes in a Multi-scale Ocean-coupled Non-hydrostatic (HMON) model, will be presented. Results from the testing of a variable consensus will also be explored.
DISCLAIMER: The views, opinions, and findings in this article are those of the authors and should not be construed as an official National Oceanic and Atmospheric Administration (NOAA) and/or U.S. Government position, policy, or decision.