Verification of tropical cyclone genesis prediction in a suite of operational global numerical weather prediction models

The first step in developing a forecast tool is to define what is being forecast and how it will be verified. While many displays of model forecasts of existing or forming tropical cyclone (TC) tracks exist, much less attention has been given to issue of how model storms compared to both existing 'numbered' TCs and 'invest' systems that are the precursor to all TCs or what I call 'pTCs.'

A suite of six operational global numerical weather prediction (NWP) are evaluated for the northern Hemisphere TC season during the years 2009-2013, namely: 1) NCEP Global Forecast System (GFS); 2) ECMWF HiRESolution (HRES); 3) CMC Global Deterministic Prediction System; 4) FNMOC NAVal Global Environmental Model (NAVGEM); 5) UKMO Unified Model; and 6) ESRL Finite-volume, flow-following, Icosahedral Model (FIM). The GFDL TC genesis tracker is run to output TC tracks from TCs in the initial conditions and TC system that form during the integration. In addition to TC tracks, I also calculate basin-wide, over-ocean precipitation both large-scale and convective to relate to under-over TC formation in the model.

Genesis is defined as time of the first advisory/warning issued by the operational US TC forecast agencies JTWC and NHC as these centers are required to issue advisory/warnings for all systems that are analyzed to be tropical cyclones regardless of initial intensity. Comparison of both types of model tracks to both TCs and pTCs is rather complex but essential to defining success and an output that has operational forecast utility.

Generally we find the models in the later part of the 2009-2013 period the models do forecast genesis well, but have differing degrees of forming spurious systems ('spuricanes') or false alarms. The false alarm rate is positively correlated to the total basin-wide precipitation and negatively correlated to the ratio of convective to total precipitation (more resolved rain, more spuricanes).

Finally, products are proposed that support the 2- and 5-d NHC formation forecasts.