Skip to end of metadata
Go to start of metadata

 Status: Finalised Material from: Linus, David L., Esti, Hao, Kristian, Mohamed


1. Impact

Hurricane Florence made landfall in  North Carolina on 14 September as a category 1 hurricane. The cyclone brought torrential rainfall as it became quasi-stationary for a few days. At least 26 people was killed due to the cyclones.

2. Description of the event

3. Predictability


3.1 Data assimilation

The plot below shows the evolution of central pressure in Best Track (dots) and LWDA analysis (solid line). The analysis had difficulties to capture the intense phase on 5 September, but captured the second intensification.

Track of FLORENCE with estimated central pressure

The plots below show time-series observation statistics for temperature observations from dropsondes within 1 (left) and 5 (right) degrees of Florence. Dropsondes wind and surface pressure statistics are plotted in the lower figures.

Below is a sequence of coverage maps showing the usage of wind from dropsondes

Throughout the lifetime of FLORENCE the data assimilation seem to  consistently reduce the position errors. 

During the lifetime of Florence an increased number of observations of ocean floats was noted (see first plot).  These additional  floats were ALAMO (Air-Launched Autonomous Micro Observer) float, 10 floats deployed by WHOI in order to monitor Hurricane Florence, in the predicted path of Hurricane. These floats descend to 300m every 2 hours. The behaviour of the data assimilation was checked and  most response in our ocean system is already visible in the model background, which did a very good job in representing the cooling/deepening effect during Florence event. And DA only slightly corrected the temperature fields.

3.2 HRES

Total precipitation accumulated from HRES during the whole episode of Hurricane Florence. Base time
14/09/2018 at 00 UTC.

3.3 ENS

The plots below show the tracks (ensemble -grey, best track - black), position and intensity on 14 September 00UTC (ensemble - squares, best track - hourglass) in forecasts from 13 September (first plot) to 4 September (last plot). Early forecasts had problems to capture the northward "kink" on 7 September. The forecast from 4 September also had a majority of members turning northward over the Atlantic. This changed on the 5 September, and the sensitivity analysis further down is targeting this change. 

The plots below show the 4 most westerly (red) and easterly (green) in  the forecast from 4 September 12UTC (first plot), and the normalised difference in z500 between the cluster mean of westerly minus easterly members for 24h, 48h, 72h and 96h. The main sensitivity at 24 and 48 hours are in the easterly extent of the subtropical anticyclone. The anticyclone was weaker in the members going towards U.S.  

The plots below shows EFI and SOT for 3-day accumulated precipitation valid 14-16 September. In the earliest forecast presented here (from 8 September), the SOT is already high while EFI is low. This is a sign of a few extreme members in the ensemble. As the forecast lead time decreases, the EFI increases as well.

The plot below shows the evolution of forecasts for 3-day accumulated precipitation (14-16 September) for Wilmington, North Carolina (34.3N,78W). The plot includes ENS (blue box-and-whisker), HRES (red dot) and model climate (red box and whisker). Although a few members highlighted the risk of extreme rainfall before, the majority of the ensemble picked up the signal gradually from 10 September 18UTC and onwards.

3.4 Monthly forecasts

The plots below show weekly strike probability (left) and normalised accumulated cyclone energy from extended-range forecasts, valid for 10-16 September. During the same week as Florence was active, TC Helene was present in the central Atlantic and TC Isaac propagated towards the Caribbean.  The earlier part of the hurricane season had been quite, and the extended-range forecast seems to picked up the signal of a more active period 2-3 weeks before.

3.5 Comparison with other centres


The predicted hydrological impacts of Hurricane Florence and is shown in the  GLOFAS forecasts below from 14 September. The rainfall is plotted over the top of the map and the hydrographs highlight the time when the highest river flow is expected.

From the Flood Inundation Mapping and Alert Network (FIMAN) website ( from the North Caroline government, you can see the current state of the rivers in real time and compare them with past forecast. As seen above, there was several warnings in the area with possible flooding and they are reflected in the images below. Some of the gages had at this point already reached the maximum level for the vent. Another interesting point from this is that , at least from the predictions of their hydrological model, it seems that at least 3 separate rivers in 3 different states will all be near or the record flood levels (see the three river flow plots). 

Map of flood warnings and current state of the gauges from the Flood Inundation Mapping and Alert Network (FIMAN) valid for today at 14:00 UTC.
Forecast and observations of the flux from three different river gauges in Manchester, Cheraw and Danville (USA)

4. Experience from general performance/other cases

5. Good and bad aspects of the forecasts for the event

  • Issues with mid-Atlantic "knee" and intensification

6. Additional material


Total precipitation observed in 7 days from the hurricane Florence.
Plot of the maximum rainfall caused by North Atlantic and Northeast tropical cyclones and their remnants per state (1950-2018).