Dominica
Using the OpenLISEM model, detailed multi-hazard simulations for the impact of hurricane Maria on Dominica were carried out. Dominica is a small carribean island (~750 km2) with a population of around 70.000 poeple. The vulcanic slopes are draped with rich tropical forests and mixed tropical agriculture such as sugar cane and bananas.
In the past decades, several high-category hurricanes have hit the island. Most notably, Hurricane David (1972), Hurricane Erika (2012), and finally Hurricane Maria (2017). On September 19, 2017, the category 5 hurricane named Maria made landfall on the Island of Dominica. Just before making landfall, there was a short and intense period of intensity growth, unpredicted by local and regional forecasters. The hurricane lead to extreme wind speeds, storm surges and cumulative rainfall throughout the island. Damages where caused to virtually all buildings, both due to winds damaging the roofs or walls, or floods and landslides damaging the buildings. Over 70 percent of all buildings where reported to have been significantly damaged during the event. Hurricane Maria lead to a total of 65 fatalities (Dominica News Online, 2018). The total damage was assessed to be 1.37 billion USD (ReliefWeb, 2018), over twice the GDP of the island.
The elevation of the island, as well as the impact inventory (featuring debris flows, landslides and flash flood processes) are shown below.
The precipitation patterns for hurricane Maria are shown below, from ground-based radar data from Guadaloupe and Martinique. The total precipitation highlights the extremity of the event, with in most locations a total rainfall of over 0.5 meters in a single day.
Below, you can find a comparison of flash flood simulations of a full physically-based model (Saint Venant Equations), and the FastFlood model as provided on www.fastflood.org.
These simulations take less than 2 seconds, a 1800x speedup compared to full traditional modelling.
In addition, we can compare mapped flood extent (hand-mapped from high-resolution sattelite imagery) with our model results and calibrate to compare the performence of the models.
Finally, the accuracy of the FastFlood method reached 95.54% compared to 95.69% for full Saint-Venant dynamic simulation.
In addition, we see incredible speed increases when doing flood simulations at a larger scale.
Below, you can see the results for the complete island of St. Lucia.
For more detailed comparisons, please see our scientific article.