Potential Vorticity Diagnosis of Hurricane Track Forecasts in IFS, GFS,
and GFDL SHiELD models
Abstract
In this study, we used the potential vorticity (PV) diagnosis technique
to investigate the key factors that affect the track forecasts of
Hurricane Maria (2017) in the NCEP GFS v14, ECMWF IFS and GFDL SHiELD
models. In Chen et al. (2019), it showed that a slow bias of Maria’s
translation speed in the IFS 5-day forecasts was significantly improved
by GFDL SHiELD with IFS initial conditions (SHiELD_IFS). Our results
found that the slow moving bias in the IFS is mainly due to a strong,
westerly steering flow contribution from a cutoff low from the northeast
quadrant and another low system from the southwest quadrant of Maria. On
the other hand, the SHiELD_IFS improves on the IFS by better simulating
the strength of the Bermuda High, and low systems in the southwest,
northwest, and northeast quadrants allowing for better track alignment
with observations. We also found that the northward track bias of Maria
in the legacy GFS and SHiELD with the GFS initial conditions
(SHiELD_GFS) was associated with a weaker Continental High which
contributed less northerly steering flow compared to that in the IFS and
SHiELD_IFS. Furthermore, the Bermuda High was relatively weak in the
SHiELD_GFS, while the two low systems in the northwest and northeast
quadrants contributed steering flow opposing Maria’s moving direction,
causing a slowdown of translation speed of Maria in the SHiELD_GFS. By
performing this piecewise potential vorticity diagnosis on all of the
storms in the 2017 North Atlantic Hurricane Season, we could possibly
identify the key elements that generate the biases in TC track forecasts
in these models.