Navigating the Storm

Abstract

Tropical cyclones, known as hurricanes in the Atlantic and Northeast Pacific or typhoons in the Northwest Pacific, are intense storm systems characterized by strong rotating winds, heavy rainfall, and low atmospheric pressure. They form over warm tropical waters and are major drivers of coastal flooding in tropical and subtropical regions. Annually, around 50 cyclones reach hurricane strength, causing flooding through storm surges and heavy rainfall, threatening communities and ecosystems. Climate change and human activities exacerbate these risks. Accurately predicting coastal flooding due to these cyclones is challenging due to their complex features, limited historical data, and forecasting uncertainties.

This dissertation aims to enhance the reliability of coastal flood forecasts and risk analysis by improving the descriptions of tropical cyclone wind geometry and pathways. It addresses both operational (short-term) and strategic (long-term) flood risk analyses. Operational risk analysis involves forecasting days before and after a cyclone, while strategic analysis deals with climate variability over decades. Both are crucial for comprehensive climate risk management, offering different time frames for preparedness and prevention.

A key element in both types of analyses is accurately representing tropical cyclone conditions in computational models. By examining historical best-track data, empirical relationships for two tropical cyclone geometry parameters—the radius of maximum winds and the radius of gale-force winds—were derived, improving estimates by up to 25%. This improvement is significant, particularly for cyclones outside the United States. These parameters, either observed or derived, are essential for computing surface wind distributions using the Holland wind model, which is critical for coastal flood evaluations.

Strategic risk analyses often suffer from a lack of sufficient historical tracks for reliable flood hazard assessment. To address this, an empirical track model based on Markov chains was introduced, capable of simulating thousands of synthetic storm pathways. The Tropical Cyclone Wind Statistical Estimation Tool (TCWiSE) generates these tracks, showing good agreement with historical data and extreme wind speeds. This methodology enhances the estimation of extreme cyclone conditions for strategic risk analysis.

The combined data-driven and physics-based methods were used to quantify coastal flooding in the Southeast Atlantic Coastal Zone of the United States. Comparing cyclone-induced flooding to non-cyclonic flooding revealed that extratropical cyclones are responsible for frequent flooding, while tropical cyclones cause the majority of infrequent but severe floods. For example, with current sea levels, extratropical cyclones contributed to half the flooded area, but tropical cyclones accounted for ~96% of the flooded area for 100-year events, affecting significantly more people. At higher sea levels, tropical cyclone-specific flood risk diminished as areas became uniformly susceptible to flooding. This analysis highlights the importance of considering both cyclone and non-cyclone flood factors in future research.

Operational risk assessments, critical for protecting lives and minimizing economic impacts, involve simulating numerous ensemble members to account for uncertainties in cyclone track, speed, and intensity. The Tropical Cyclone Forecasting Framework (TC-FF) integrates major physical drivers such as tide, surge, and rainfall, using Gaussian error distributions and autoregressive techniques. A case study of Cyclone Idai in Mozambique demonstrated the need for a large number of ensemble members for reliable forecasts. TC-FF showed less than 10% difference from operational ensembles, suggesting its utility in data-scarce environments.

This dissertation provides new insights into tropical cyclone wind geometry, pathways, and their role in compound flooding. Future research should enhance data collection, particularly from satellites, to validate models and understand storm characteristics better. Incorporating overlooked processes, leveraging data assimilation, and exploring more efficient methods, including Deep Learning, are essential for advancing flood risk assessment and capturing tropical cyclone variability.