Simulating the Southwestward Looping Track and Rapid Intensification of Hurricane Joaquin (2015)

Hurricane Joaquin (2015) was an unusual case with a southwestward track and rapid intensification. It originated as an extratropical mid-to-upper level low that tracked southwestward toward the Bahamas and acquired a surface reflection with associated deep convection, while remaining in a moderately sheared environment. Designated as a tropical depression at 0000 UTC 28 Sep, the storm slowly intensified until 0600 UTC 29 Sep, when it began a 60-h rapid intensification aided by high SSTs, with its maximum surface wind increasing from 35 to 115 kts. The normally reliable GFS and UKMET models held steady with forecasting the storm to curve westward, then northwestward, and make landfall on the mid-Atlantic coast through the 1800 UTC 30 Sep forecast cycle, when the storm was within 4 days of a potential U.S. landfall. In contrast, the ECMWF model had several days prior caught onto the idea of the storm tracking further southwestward into the Bahamas, before sharply recurving to the Northeast, missing the U.S. coast, which turned out to be the correct solution.

Here, we present the results of our 1-km Weather Research and Forecasting (WRF) model prediction, which successfully simulates the full 60-h rapid intensification period and looping southwestward track. We also show verification of our simulated storm structure against various in situ and satellite observations, including those obtained during the Tropical Cyclone Intensity (TCI) field program. To generate realistic synoptic-scale and vortex-scale initial conditions for the deterministic simulation, we run a 24-h dynamical WRF spinup experiment with a 3-h cycling period, in which conventional, radiance, high-resolution AMV, and aircraft reconnaissance mission observations are assimilated into a 3-km background forecast using the WRF Data Assimilation system (WRFDA) in hybrid mode. Ensemble error statistics are derived from an 80-member 9-km WRF ensemble updated using a similar set of observations and the Data Assimilation Research Testbed (DART) EnKF.

Through a series of sensitivity tests, we show that supplying the model accurate initial conditions on the vortex scale, using cycled data assimilation, was essential in capturing the initial southwestward motion and onset of rapid intensification. We also examine the complex synoptic-scale steering environment surrounding the rapidly intensifying Joaquin and compare our model-predicted synoptic scale features against the operational GFS analyzed features. Using this comparison, along with a few sensitivity tests and analysis of the DART ensemble spread, we attempt to better understand 1) the key synoptic-scale features responsible for Joaquin’s unusual track and 2) possible reasons why the operational GFS model struggled with this storm’s track prediction.