New Challenges and Expectations of Dynamical Seasonal Prediction of Tropical Cyclones

Seasonal prediction of hurricane activities in the north Atlantic basin has been shown to be remarkably skillful during 2001-2010 (Chen and Lin 2011). However, it has met its challenges since 2011. For the 2013 season, Atlantic hurricane activity was much quieter than most forecasting groups predicted. For the 2011 and 2012 seasons, many weak TCs formed, verifying TC number predictions, but only a few of them intensified into major hurricanes, a bust hurricane number prediction. After Hurricane Sandy's landfall near New Jersey and New York metropolitan area in 2012, the more challenging problem of assessing the probability of landfall events in a seasonal prediction has started to draw attention.

To address the above issues, a new version of Geophysical Fluid Dynamics Laboratory (GFDL) High-Resolution Atmospheric Model (HiRAM) is being developed. In Chen and Lin (2011, 2013), the 25-km GFDL HiRAM showed high prediction skill of Atlantic tropical storm counts in the 21 years since 1990, with a correlation of 0.94 between the observed and model predicted hurricane counts during 2000-2010. In the new generation of HiRAM, a non-hydrostatic option has been added, with improved microphysics and convection schemes, making the model more suitable for cloud-resolving simulations. With the finer vertical resolution and higher model top, the improved modeling system can now better resolve the upper troposphere and the stratosphere, thus improving the simulation of the tropical quasi-biennial oscillation (QBO) and Madden-Julian Oscillation (MJO), which may play important roles in the seasonal TC activities. The forecast skill of the new HiRAM for the past decade and recent seasons, using the persistent sea surface temperature (SST) assumption, will be presented.