Introduction
A tropical cyclone forecast involves the prediction of several interrelated features, including the track, winds, rainfall and storm surge. Many organisations provide forecasts and for example the National Hurricane Center (NHC) in Miami, FL issues a forecast every 6 hours from 72 hours till landfall. However, tropical cyclone forecasts are not perfect. Errors arise from a lack of a full understanding of the formation and growth of tropical cyclones and from the limitations of the forecasting techniques themselves. For these reasons it is important to take into account some variability in the track and intensity of the storm when modelling cyclones with Delft3D-FLOW and WAVE. This is done with so-called 'Ensemble modeling'. Ensembles for cyclones can be created with the Advanced Tropical Cyclone.
Background of the toolbox
The first two tabs of the Advanced Tropical Cyclone toolbox are similar compared to the Tropical Cyclone toolbox. A third tab is added called 'Ensemble'. Within this tab we are going to create multiple spiderwebs of a cyclone forecast, but this time with taking into account the uncertainty related to the prediction. The idea of the Advanced Cyclone Toolbox follows De Maria et al. (2009). De Maria et al.(2009) states that the track error is the great-circle distance from the forecast to the best-track position. The error vector is decomposed into along-track (AT) and cross-track (CT) components, relative to the direction of the cyclone motion vector in the forecast track. The ATE and CTE are determined by a simple autoregressive technique. For the error in the maximum wind (VE) something similar is done, but De Maria et al. (2009) also takes into account distance to landfall. In Delft Dashboard these relationships have been simplified somewhat and only the linear function of the error at the previous time period is taken into account
The along-track track error (ATE) and cross-track error (CTE) are in Delft Dashboard inserted in meters per 12 hours. One can apply for example the forecast error from NOAA here. These forecasts errors are however provided as great-circle distance in nmi (nautical miles) for a certain time period before landfall. The user of the Advanced Tropical Cyclone toolbox should determine how this is divided into an ATE and CTE (e.g. an equal distribution and divide the great-circle distance with square-root of 2). On top of that, the error should be given in 12 hour periods. A similar approach should be applied for the error in maximum wind. The VE is inserted in kt (knots) per 12 hours.
Usage of the toolbox: example of how to determine which errors to apply
In this example we are going to model the hazards of a cyclone hitting on New York City. We are 48 hours before landfall and would like to have an idea of the water levels and wave heights at the coast of New York. This modelling can be done in 5 steps:
- Generate a Delft3D-FLOW and WAVE model. In this case we can use a model developed for Hurricane Sandy.
- Import the track of the storm (forecast).
- Look up the average errors in forecast track (ATE and CTE) and intesity (VE) in the basin your looking (e.g. Atlantic, Indian or Pacific)
- The average error in the track 48 hours prior to landfall is 77 nmi. This is 142.450 meter for 48 hours and 35.613 meters for 12 hours. Divided with square-root of 2 this results in a 25.181 meters ATE and CTE. It is also possible to only focus on the ATE error or vise-versa for the CTE.
- The average error in intensity 48 hours prior to landfall is 11.4 knt. This is 2.85 kt per 12 hours.
- Generate the full ensemble with a number of realizations. Create a 1.000 till 10.000 tracks.
- Make spiderwebs for a limited ensemble with the variation in cross-tracks (ncross), forward speed (spd) and maximum wind speeds (nvmax). The full ensemble is used to determine the probability of occurrence.
Figure: example of ensemble forecast for the Atlantic Ocean. In this example, we used the track of Hurricane Sandy
