Paper on what led to the very uncertain forecasts of Hurricane Joaquin in 2015 published in Weather and Forecasting


During the 2015 North Atlantic hurricane season, Hurricane Joaquin meandered near the Bahamas.  Forecasts ranged from bringing Joaquin inland over the U.S. to it going east of Bermuda, suggesting that Joaquin’s track was very uncertain (Fig. 1).  A high-resolution forecast model based on NOAA’s hurricane prediction system, the Hurricane Weather Research and Forecast (HWRF) model, was used to study this large uncertainty.

In general, hurricane track forecasts depend on what is happening in the air surrounding the storm (what we call the environment) and what is happening in the storm itself. Hurricanes are mainly steered by the winds associated with high- and low-pressure weather systems in the environment, what we call the steering winds. Characteristics of the hurricane are important for track, as well, because taller, more intense hurricanes may feel different steering winds higher up than shorter, weaker ones. Because we cannot measure these steering winds or storm characteristics everywhere, we can’t know exactly what is happening. To account for this, we can run the model many times from different starting points (called initial conditions) that fit the observations we do have.  This set of model runs is called an ensemble.  We created an ensemble of eighty forecasts and investigated why some made landfall in the U.S. and why others did not. Specifically, we looked at the steering winds and what was happening inside Joaquin itself, things such as intensity, structure, and location, to see what made the forecast so uncertain.

To test if differences in the initial location of Joaquin was important to track forecasts, a new technique was developed to reposition Joaquin in the initial conditions.  This technique can be used easily in other models for research projects.

Screen Shot 2019-09-10 at 10.33.11 AM.png
Figure 1. Eighty track forecasts from an HWRF-based ensemble for Hurricane Joaquin. The thick black line shows Joaquin’s actual track.


  • Important Conclusions:  
  1. Track forecasts are highly depended on steering wind far above the earth’s surface, especially weak steering winds near the Bahamas (red in Fig. 2). These steering winds were controlled by three weather systems: low pressure over North America (green), high pressure in the central North Atlantic (blue), and low pressure to Joaquin’s east (magenta). Accurate model initial conditions of these weather systems were needed for accurate track forecasts. 
  2. Track forecasts were more accurate if Joaquin became embedded in the weak steering winds between these three weather systems. On the other hand, track forecasts predicted U.S. landfall if these weak steering winds were located east of Joaquin.
  3. Unexpectedly, Joaquin’s intensity, structure, and initial position were not important to track forecast uncertainty. Even when Joaquin’s initial position was artificially moved hundreds of kilometers, the changes had little impact on track forecasts.
    Screen Shot 2019-09-10 at 10.33.58 AM.png
    Satellite image from October 2, 2015, showing Hurricane Joaquin and the different features that impacted its track.

    You can read the full article at

    Screen Shot 2020-02-19 at 4.37.41 PM.png