Examination of GFS Medium-Range Tropical Cyclone Track Forecasts in the North Atlantic (2012–2014)

The aim of this presentation is to examine and diagnose medium-range tropical cyclone (TC) track forecasts from the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) over the North Atlantic basin for 2012–2014. This study is motivated by the medium-range GFS track forecast statistics for this period, which show a significant increase in mean absolute track error (MAE) for forecast leads > 4 days relative to the Hurricane Forecast Improvement Program (HFIP) “baseline” track forecast error established using the top-flight numerical models for the 2006–2008 TC seasons. Close inspection of the distribution of absolute error at day 7 reveals that while nearly half of all 7-day forecasts were improved relative to the HFIP baseline, the positive skewness of the distribution brings the MAE above the HFIP baseline. Clearly, a relatively small number of extremely poor track forecasts (or, “busts”) are negatively influencing the overall medium-range TC track forecast statistics.

We classify the GFS 7-day track forecast busts (with busts defined as absolute track error 1.5+ sigma larger than the MAE for 2012–2014) based on the synoptic-scale flow. For the 25 busts identified out of 287 forecasts, the classifications are as follows (with counts in parentheses): “enhanced recurvature” (13), “trough missed” (4), “missed recurvature” (2), “ET life cycle” (2), and “unclassifiable” (4). The “enhanced recurvature” classification is characterized by TC track forecasts that recurve the TC into midlatitudes too early or sharply. The enhanced recurvature category is comprised of five forecasts for TC Sandy (2012), five for Isaac (2012), two for Arthur (2014), and one for Humberto (2013). The forecasts for Sandy, Arthur, and Humberto are marked by an upstream trough along the subtropical/midlatitude waveguide with reduced meridional amplitude. The increased westerly flow that accompanies the more zonal trough/waveguide rapidly carries the TCs out to sea. The reduced amplitude of troughs and ridges is fairly persistent for all GFS medium-range forecasts for 2012–2014. The GFS forecasts for Isaac are a bit more complex, as the inability of the GFS to produce a precursor mesoscale convective system ahead of Isaac led to significant errors in the synoptic-scale flow. Additionally, in this presentation we will compare the enhanced recurvature forecast busts with a subset of the best GFS forecasts in this synoptic-scale flow classification for 2012–2014. We aim to assess if any key differences are apparent in the synoptic-scale precursor or in the TC structure itself that makes a forecast “bust” more or less likely.