2012 Upgrades to the Operational GFDL/GFDN Hurricane Model

The operational versions of the GFDL model used by the National Weather and the US Navy (GFDN) were upgraded in 2011 with the new version of the Simplified Arakawa-Schubert (SAS) deep convection scheme operational in NCEP's global model (GFS). Tests from the 2010 Atlantic hurricane season demonstrated reduction in track error at 4-5 days of nearly 20% with the new deep convection. Despite some degradation in intensity performance, particularly for weak storms, the upgraded model was made operational in 2011. During the 2011 season evaluation of the model performance, indicated an excessive west track bias with the upgraded model, partiuclarly for Hurricane Irene. Much of the bias was traced to the lack of detrained micro-physics in the new SAS scheme. With this corrected made, and the micro-physics generated and properly passed from the SAS scheme to the micro-physics scheme, the track errrors were signficantly impacted, in both the East Pacific and Atlantic. A similar result was found in the GFDN model for the parallel runs made in the Western Pacific. Further investigation indicated the positive intensity bias could be significantly reduced by increasing the cumulus momentum tranpsort term. These changes are being extensively evaluated for likely implementation in th 2012 versions of the GFDL/GFDN models. Additional upgrades that are aslo being testing for the 2012 upgrade, include implemenation of the GFS shallow convection and new planatary boundary schemes, and upgrades to the GFDL model's Ferrer micro-physics. Preliminary testing of these changes indicates significant overall improvements in the intensity prediction, and reduction of track error particularly in the East Pacific, and some cases in the Atlantic. One of the largest improvements in track was found in Hurricane Irene, with the average 3 and 5 day track error reduced from 152 and 360 nm, to 90 and 149 nm, respectively. In this talk, the proposed changes will be discussed in detail, the final model configuration will be presented, as well as evaulation of the new model's performance for track and intensity, for reruns of the 2010 and 2011 Atlantic and East Pacific hurricane seasons. Finally, imporvements to the model initialization planned in future upgrades will be discussed.