On 27 August 2020, Hurricane Laura made landfall in southwestern Louisiana as a category 4 hurricane, making it one of the most intense cyclones to make U.S. landfall based on maximum sustained wind speed. The effects of the hurricane were a damaging storm surge, extreme winds, and coastal and inland flooding. However, the impacts could have been much larger if the cyclone had hit the metropolitan area of Houston, Texas, with approximately seven million residents and large oil and petrochemical industries. Prior to its intensification into a hurricane, Tropical Storm Laura caused substantial damage in the Caribbean, especially Haiti, the Dominican Republic, and Cuba, including several dozen fatalities.
The tropical cyclone originated from an African Easterly Wave that propagated from western Africa into the Atlantic around 15 August. At this stage ECMWF’s ensemble forecast predicted the risk for the genesis of the cyclone, but with high uncertainties about the future path and duration of the system. For example, the ensemble from 17 August included a few members with a cyclone entering the Gulf of Mexico, but also members hitting Florida and curving north over the Atlantic, and with many members where the cyclone quickly dissipated.
Predictions in the Gulf of Mexico
The cyclone was recognised as a tropical storm on 21 August while being east of the Leeward islands. The cyclone passed over land on Hispaniola and Cuba and entered the Gulf of Mexico on 25 August as a weak cyclone. At this point the ensemble predicted a landfall on the north-western coast of the Gulf on 27 August, but with large uncertainties. This is a region with sharp gradients in the population, and a relatively small difference in the landfall position would result in huge differences in the human impact. In the forecast from 00 UTC 25 August, the city of Houston, Texas, was in the middle of the ensemble plume. This was a critical time for local officials to decide whether or not to issue a mass evacuation order for the city. This remains a sensitive issue given the chaotic evacuation ahead of Hurricane Rita in 2005, which resulted in over 100 fatalities. On the morning of 25 August 2020, the United States National Hurricane Center reviewed all the latest operational numerical model and ensemble forecasts, and they decided to nudge the forecast landfall location slightly nearer to Houston than their previous forecast, but not directly at Houston. As a result, no mandatory evacuation order was issued for greater Houston, with only the local barrier islands being evacuated at that time. On 26 August, the ensemble plume shifted further to the east and eventually the hurricane made landfall on the coast of Louisiana, with Houston receiving only minor impacts.
It was noted by many users that on 24–25 August the ECMWF high-resolution (HRES) forecast was located on the eastern edge of the ensemble plume, which posed the question if the lower resolution of the ensemble degraded the result. To test this hypothesis, subsequent experimental ensemble forecasts with the same resolution as HRES were run for the forecasts from 23 August 00 UTC to 25 August 12 UTC. These experimental ensembles were indeed more centred around the HRES forecast and resulted in substantially improved forecasts for most dates, with a systematic shift of the plumes to the east. However, still most members of the experimental ensemble from 25 August (both 00 and 12 UTC) had landfall to the west of the eventually observed landfall point. The exception regarding improvement was the forecast from 23 August 12 UTC, where the eastward shift led to larger errors in the sense of the ensemble mean.
Intensity forecast
The experimental ensemble also captured the intensity forecast far better than the current operational ensemble due to the higher model resolution. The ensemble gave a good indication of the rapid intensification of the cyclone over the Gulf of Mexico on 26 August. The maximum wind speed in the cyclone was reasonably captured, a result of the model upgrade in Integrated Forecasting System Cycle 47r1, reported in the Spring 2020 issue of the ECMWF Newsletter.
Open dataset and summary
The challenging forecasts for this case pose several research questions. To facilitate further research on the details of the ensemble forecasts, ECMWF has archived the full vertical resolution for the operational ensemble forecasts for TC Laura. This dataset would make it possible to do detailed studies of the tropical cyclone structure in the ensemble members, and also to downscale the system with limited-area models for detailed impact studies. The plan is to make this dataset available for the research community to facilitate further research.
