Status:Finalised Material from: Linus, Mohammed, Fernando, Ervin, Ivan
1. Impact
The hurricane Matthew hit south-western Haiti on 4 October and killed more than 900 people. The hurricane later hit Barbados (6 October) and stoke along the coast of Florida on 7 October and South Carolina on 7 and 8 October, killing at least 16 people.
https://en.wikipedia.org/wiki/Hurricane_Matthew
http://www.bbc.co.uk/news/world-latin-america-37570409
http://www.bbc.co.uk/news/world-latin-america-37602431
2. Description of the event
The series of plots below show analyses of MSLP (contour), 200 hPa wind speed (shade) and area with more than 5 mm/6h (hatched). The tropical cyclone formed north of Venezuela on 28 October. The cyclone originated from a westward propagating tropical wave that on 22 October was located on the coast of Africa. After formation the cyclone propagated westward and around 2 October the cyclone made northward turn and made landfall on south-western Haiti on 4 October. After passing Barbados the cyclone swept along the coast of Florida and South Carolina on 7 and 8 October and started to connect to the jet-stream over U.S on 9 October. On 5 October, TC Nicole formed east of Matthew.
For the forecast performance of the position, three critical moments have been identified: The northward turn (1), the possible landfall on Florida (2) and whether the cyclone was going to turn back south, which did not happened (3).
The plot below shows the phase diagram from Hart (2003), from the site http://moe.met.fsu.edu/cyclonephase/
3. Predictability
3.1 Data assimilation
The plot below show the position and central pressure from the o-suite (LWDA) analysis (left) and the evolution of the minimum pressure in best track (black dots), o-suite (black), FD e-suite (blue), RD e-suite (red) and RD experiment without drop sondes (green).
The plots below show the time-series of observation statistics for dropsondes with 1 degree (top) and 5 degrees (bottom) from the centre of the cyclone. The left plots are for o-suite and the right for FD e-suite.
The plots below show the observation statistics for temperature (left) and wind vector (middle) from dropsondes and surface pressure observations (right).
29 September 00z
29 September 12z
30 September 00z
30 September 12z
1 October 00z
1 October 12z
2 October 00z
2 October 12z
3.2 HRES
The 3 plots below shows the 24-hour precipitation for NEXRAD (top-left) and HRES (1-day: top-right, 5-day: bottom-left), for the days when the cyclone affected U.S.
3.3 ENS
The plots below show the forecast of tropical cyclone activity for for three different time periods. The first (verifying 3-5 October) is to verify the capture of the northward turn), the second (7-9 October) is for the landfall on Florida and the third is for the possible southward turn (11-13 October). For each case, all forecasts are one day apart starting from the forecast issued closest to the event.
3-5 October
The plots below shows the tropical cyclone tracks from 29 September 00z for ECMWF (left) and CMC and NCEP (right). The ECMWF ensemble showed two clusters (easterly, westerly) among the tracks. The Canadian ensemble favoured the west cluster while the NCEP ensemble only had members in the western cluster.
7-9 October
The plot below compares different centres from 3 October 00z.
11-13 October
The plot compares the forecasts from 8 October 00z and9 October 12z, indicating the shift in the forecast.
The plots below compares different centres for the forecast from8 October 00z.
3.4 Monthly forecasts
The plots below shows the tropical cyclone activity for the week of 3-10 October.
3.5 Comparison with other centres
See plot in Section 3.3
4. Experience from general performance/other cases
- For issues with drop sondes, compare with Joaquin (2015) and Gonzalo (2014).
5. Good and bad aspects of the forecasts for the event
- The cyclongensis captured from around 22-23 September, but too far west and predicted propagation into Gulf of Mexico
- Good prediction of the northward turn, but with large uncertainties (not necessary at bad thing)
- Still large uncertainties about the landfall on Florida in short-range forecasts
- Falsely predicted a southward turn in forecasts around 7 October
- Problems to assimilate drop sondes, especially for e-suite, resulting in destructed cyclone in the analyses