TC genesis was identified for all basins using the public elease of the GFDL Vortex Tracker v3.9a supported by the DTC (Biswas et al. 2018). The script and fix files used for running the genesis tracker can be found in a GitHub repository (refer to script loop_genesis.ksh).

The default configuration of the GFDL Vortex Tracker for genesis was applied to the 0.25° post-processed model output for the band of latitudes between 30°S and 30°N. It was noticed that the output files contained a large and unrealistic number of forecast genesis. A postprocessing step was then applied to eliminate spurious cyclogenesis by:

• Retaining only storms that lasted for at least 24-h with the maximum sustained surface wind speed 34 kt or more (i.e., consistent with the NHC Tropical Storm definition)
• Removing any model genesis forecasts whose forecast genesis time is at forecast hour 000. We assume that these are existing TCs - not forecasts of TC genesis.
• Removing any model genesis forecasts whose forecast genesis time is greater than 120h leading time. In this way, we examine the model performance in the first 5 days as most of the operational hurricane model does.

The model TC genesis events were then verified with the observed using a script adapted from Dr. Halperin (Halperin et al. 2013). The method is currently done to match as closely as possible the National Hurricane Center (NHC) Tropical Weather Outlook (TWO), a product that provides categorical and probabilistic forecasts of TC genesis. In this method, for each model genesis forecast, find all entries in the best-track that (i) have a time stamp that matches the valid time of the model genesis forecasts and (ii) have a lat/lon within 5 degrees of the model indicated TC lat/lon at the valid time of the model genesis forecast.

• If the initialization time is 0-120 h before the best-track genesis time (defined as first entry of TD or TS in the b-decks), then we have a hit.
• If the initialization time is >120 h before the best-track genesis time, then we have a false alarm (FA).
• If the initialization time is after the best-track genesis time, but still within 72 h temporal and 5° spatial tolerance, then we have a late genesis (LG).
  

Forecast genesis that does not meet the criteria for hit or late genesis is considered as false alarm.

Figure below shows the verification of the Northern Hemisphere. The number of hits, even considering late genesis, is quite low compared to the number of observed genesis for all suites. It is noted that Suite 2 and Suite 4 produce the most hits in 2016, while Suite 1 and Suite 2 produce the most hits in 2017. However, Suite 3 produces the least hits in both years. Conversely, all the suites have a large number of false alarms, with Suite 1 having the greatest number. As mentioned above, it is likely that some of the false alarms correspond to extratropical storms.

More results can be found at:
GMTB's Final Report