AOSC Seminar
November 2, 2017

Using Ensemble Prediction System Forecasts to Rapidly Assess Tropical Cyclone Impacts on Residential Communities

Richard Krupar
Center for Disaster Resilience, UMD

Tropical cyclones cause significant damage to the built and natural environments through the concomitant hazards of wind, rainfall and storm surge worldwide. The use of scenario analysis and risk models can provide an effective means of assessing how tropical cyclones can impact buildings, critical infrastructure and society. These simulations can also be used to better inform policy development and risk decision making. However, there are inherent uncertainties associated withdatasets and models used to drive the scenarios that make it extremely difficult to generate effective policies and risk decision protocols. Using Queensland, Australia, as the domain of interest, this study focuses on evaluating the utility of ensemble prediction system (EPS) forecasts to rapidly assess tropical cyclone impacts, specifically, wind and storm tide, on residential communities.


To rapidly assess tropical cyclone wind and storm tide impacts on residential communities in Queensland, Severe Tropical Cyclone Marcia (2015) was selected as the primary case study event to reconstruct, given that it was the southernmost Category 5 cyclone (on the Australian tropical cyclone intensity scale) to make landfall in Queensland based on the Bureau of Meteorology historical tropical cyclone catalogue. Then, a multi-hazard (i.e. wind and storm tide) tropical cyclone disaster impact scenario model was built using a wide range ofanalytical, statistical and heuristic hazard and vulnerability models. Finally, European Centre for Medium-Range Forecasting (ECMWF), Global Ensemble Forecasting System (GEFS) and United Kingdom Meteorological Office (UKMO) EPS forecasts were sourced for the 72 hr, 48 hr, 24 hr, and 12 hrforecast periods leading up to landfall of Marcia to drive the multi-hazard tropical cyclone disaster impact scenario model.


Results from this study reveal that EPS forecasts cannot predict rapid intensification or storm size well, which presents a major issue when reconstructing landfall events. Manual modifications to the EPS forecasts of minimum central pressure and storm size (i.e. radius of maximum wind or RMW) were required to generate realistic hazard wind and storm tide footprints. Another challenge presented was the validation of impacts on residential communities. Due to a lack of detailed information on the building stock and demographics of the threatened region, it was incredibly difficult to trust the impacts generated by the wind and storm tide vulnerability models, despite the fact that the hazard footprints were well calibrated against landfall event-specific observations. This study highlights the need for more research on the inherent limitations of EPS forecasts and the collection of more detailed exposure information to build better physical vulnerability models that can generate realistic impacts.