Hurricane Forecasts with a Mesoscale Suite of Models

A suite of mesoscale models are being used in the present study to examine experimental forecast performance for tracks and intensity of hurricanes covering the years 2004, 2005 and 2006. Fifty-eight storm cases are being considered in the present study. Most of the mesoscale models are being run at a horizontal resolution at around 9 km. This includes the WRF (two versions), MM5, HWRF, GFDL and DSHP. The performances of forecasts are evaluated using absolute errors for storm track and intensity. Our consensus forecasts utilize ensemble mean and a bias corrected ensemble mean for these member models on the mesoscale and the large-scale model suites. Comparing the forecast statistics for the mesoscale suite, the large-scale suite and the combined suite we find that the mesoscale suite provided the best track forecasts for 60 and 72 h. However, the forecast from the combined suite of model were also very close to the track errors of the mesoscale at 60 and 72 h. Overall track forecast errors were least for the combined suite. The intensity forecasts of the bias corrected ensemble mean of the mesoscale suite were comparable to DSHP and GFDL at the later part of the forecast periods.