Jeff Kepert, Stefan Zieger, Saima Aijaz, Diana Greenslade, Aaron Wassing
Experience has shown that the observational record of tropical cyclones is too short to directly estimate long return-period risk for offshore or coastal areas. Accordingly, such risk is usually estimated through the analysis of synthetic datasets.
A new synthetic Tropical Cyclone dataset representing 100,000 years for the Australian region has been developed. The key parameters are cyclone position, movement, maximum wind speed and radii of maximum winds and gale-force winds. The tracks were trained on Best Track data since 1979, but the intensity and structure parameters on post-2003 Best Track data due to inhomogeneities in the historical record earlier than this.
Gradient-level wind fields were based on the parametric wind profile of Willoughby et al (2006). These winds are adjusted for the effects of friction using the dynamical boundary-layer model of Kepert and Wang (2001). A statistical adjustment of the winds is also available, which was trained on output of the dynamical boundary-layer model.
These winds were merged into environmental fields from the ERA-I reanalysis in order to drive the wave simulations. It was desired to merge storms into synoptic situations that were consistent with the presence and track of a tropical cyclone. Accordingly, analogues to the synthetic storms were identified in the observed dataset, then the dates of the analogue were used to determine the merging fields. The reanalysis fields had the existing vortex removed to facilitate the merging.
Wave simulations were undertaken with WAVEWATCHIII® on a 5 km grid with two dedicated 1 km grids around shallow reef areas. Boundary conditions were derived to align in time with the merged fields.
The wind and wave modelling procedures were validated for about 15 historical Tropical Cyclones that occurred in the north-west Australian region.
The dynamical boundary-layer and wave models are computationally expensive, so were used only for storms that were estimated to have high impact at locations of interest by simpler models.
This presentation will give a broad overview of the methods and results and show some verifications. Challenges will also be discussed.