Reask was created to provide natural catastrophe risk analytics anywhere and everywhere on the globe, and today we are proud to release the first step of that journey: a global view of Tropical Cyclone (TC) risk.
Our Unified TC model (UTC) is designed to represent 100k years of global TC wind activity and allows computation of probabilistic hazard layers for all TC affected countries worldwide. UTC is built from:
- A global stochastic set of events, explicitly connected to climatic conditions and weather patterns
- A global wind field model designed to reproduce key regional differences in TC wind shapes and asymmetries observed across the globe
- A global terrain-adjustment model allowing simulation of local wind gusts and capturing the diversity of terrain roughness and topographic features worldwide.
Climate-connected event set
To simulate key event characteristics and allow their generalization beyond basins where historical records are abundantly available, we have trained the UTC model to understand the main climate drivers of TC activity. Using a combination of ERA51 reanalysis and ensemble members from the NCAR-CESM2 model as forcing, all synthetic events in the 100k years UTC catalogue are explicitly aware of global climate conditions and local weather patterns.
This results in a physically driven model where sea surface temperature, steering flow and wind shear patterns all play a role in simulating event characteristics such as:
- genesis frequency & location
- track trajectories
- intensification rates
It is by moving away from a purely statistical approach towards our more physically driven framework that the UTC ensures consistent risk assessment on the global scale, even in regions where data is scarce.
By being fully aware of climate variability, the UTC model is also in a unique position to quantify the impact of the main climate cycles such as ENSO (Fig. 2), and can provide refined views of risk that reflect the latest climate observations (e.g. seasonal trends).
Global wind field model
To capture the range of wind field shapes and asymmetries occurring across the globe, we have trained our Machine Learning (ML) model3 using high resolution data from InCyc: our proprietary database of global TC simulations.
The result provides a unique (non-parametric) model able to distinguish between wind field shapes from systems as varied as pure tropical hurricanes in the deep tropics and extra tropical transitioning typhoons around Japan.
For all events in the UTC catalogue we deployed this ML model to generate 1-min sustained, over water equivalent winds at 1 km horizontal resolution.
Global terrain correction model
As a final step we have trained additional ML models to convert over water sustained wind estimates into local 3-sec gusts over actual terrain. The models, trained using InCyc 1 km gust data, account for the impact of changes in terrain roughness and topography both at site, and up to 3 km upwind (i.e. directional wind correction).
This roughness & topography correction approach has been extensively tested for terrains in the US, Japan and Australia and can handle the diversity of terrain features present in these 3 countries.
Recent examples of terrain corrected footprints are available via our event response solution HindCyc (see Fig. 4 for an example from Typhoon Haishen 2020).
Global hazard layers
The end result from the UTC workflow is a global catalogue of wind gust footprints representative of 100k years of climate-connected TC activity. In addition to making the granular hazard footprints available for consumption by organizations, we also provide probabilistic hazard layers in the form of gust return period (RP) maps (e.g. Fig. 5).
1-km resolution, terrain-corrected probabilistic gust maps such as the one above are available for all TC affected countries worldwide, at any return period level.
Our Unified Tropical Cyclone solution brings together more than 2-years of intensive development to provide a view of tropical cyclone risk anywhere across the globe – please get in touch for more information.
1 Hersbach, H, Bell, B, Berrisford, P, et al. The ERA5 global reanalysis. Q J R Meteorol Soc. 2020; 1– 51. https://doi.org/10.1002/qj.3803
2Kay, J. E. et al. The Community Earth System Model (CESM) Large Ensemble Project: A Community Resource for Studying Climate Change in the Presence of Internal Climate Variability. Bull. Amer. Meteor. Soc. 96, 1333–1349 (2015).
3Loridan, T., R.P. Crompton, and E. Dubossarsky, 2017: A Machine Learning Approach to Modeling Tropical Cyclone Wind Field Uncertainty. Mon. Wea. Rev., 145, 3203–3221.