Status:Ongoing analysis Material from: Linus, Ivan
Picture
1. Impact
On 2 September the tropical cyclone Hermine made landfall on the north-western side of Florida.
http://www.bbc.co.uk/news/world-us-canada-37248359
https://en.wikipedia.org/wiki/Hurricane_Hermine
During its life-time the cyclones challenged the forecasting systems several times. Firstly, whether or not the cyclone was going to develop east of Florida, later about the landfall on the Gulf of Mexico and finally about reintensification over the Atlantic and a landfall on the U.S east-coast.
2. Description of the event
The sequence of plots below shows 12-hour forecasts of MSLP and precipitation (as a substitute of analyses) valid from 28 August 00z to 6 August 00zwith 24 hours apart. Before the sequence starts a convective system was present over the Caribbean and was on 25 August located north-west of Puerto Rico. On the 28-29 August the convection reached Florida. TC Hermine was upgraded to hurricane on 1 September and make landfall on northern Florida on 2 September. The storm later moved to north-east.
3. Predictability
3.1 Data assimilation
During different stages of the evolution of the Hermine, extensive dropsonde missions took place, e.g by using the NASA Global Hawk platform. The plots below show the observation statistics for dropsondes for temperature (left) and wind vector difference (right) for the days with most dropsondes.
25 August 12z
30 August 00z
30 August 12z
31 August 12z
1 September 00z
1 September 12z
2 September 00z
3.2 HRES
The plots below shows the HRES forecasts from 25 August and 26 August 00z for the MSLP and rainfall for 30 August 00z. The figures illustrates the the change in the forecast regarding the cyclogenesis east of Florida around 29-30 August. The change in the forecast took place at 25 August 12z.
Tropical cyclone Hermine made landfall over Florida and passed over few other states on Friday, 2nd September. It brought a lot of rain, strong winds and caused flooding. Short-range HRES provided a good precipitation forecasts. The shape of the rain band was very well captured in the forecast compared to the NEXRAD.
3.3 ENS
The plots below show the tropical cyclone activity for 31 August to 1 September. Already in the earliest forecast (from 24 August) the probability for a tropical cyclone in the northern part of Gulf of Mexico was above 20% and in the forecast from 25 the probability increased to above 50%. However, in the subsequent forecasts the probability decreased again. This change was connected to the whether the cyclone should form already in the Atlantic, which it did not. But later Hermine formed in the same tropical depression.
The intensification of Hermine that was in early forecast once it starts moving over the open sea did not materialize. When Hermine made landfall on Friday, 2 September, it lost a lot of its intensity and more or less it's kept it later when it moved again over the ocean whilst some early forecasts suggested intensification especially the HRES. The good point is that the reported core pressure falls within the ENS range although at the upper tale of ENS distribution. It's also worth mentioning that the ensemble spread was quite large and the HRES was also within the range of the ensemble but at the lower tail. Regarding the track, Hermine moved north-eastwards at some point it changed the direction approaching the north-east US coast. That turn in thr direction was given by a number of ensemble members as well.
Verification of Hermine's track forecast based on 2 September 00UTC run. Red dots represent the reported core pressure of Hermine.
3.4 Monthly forecasts
3.5 Comparison with other centres
Two other global prediction centres, UKMO and NCEP, also provided good short-range forecasts. It's worth mentioning that the resolution of HRES is high enough to capture structures in the rain field which are more realistic than lower resolution global models. In the very short-range forecast (T+12-36h) we can spot the eye of the cyclone with less rain than the surrounding area and spiral structures that were also observed although with some location errors.
4. Experience from general performance/other cases