Uncertainty Quantification for a Climatology of the Frequency and Spatial Distribution of North Atlantic Tropical Cyclone Landfalls

Abstract:

We develop a Bayesian hierarchical model for the climatological frequency of Atlantic Basin tropical cyclone (TC) landfalls along the coast of North and Central America. The model is explicitly spatial, with a covariance structure that incorporates the effects of coastline geometry, and is resolved at impacts-relevant, 50-mile coastal increments. The model is based on a negative binomial regression on the phase of the Southern Oscillation, North Atlantic Oscillation, and the Atlantic Multidecadal Oscillation, and also accounts explicitly for the time-dependent uncertainty in the historical data used to fit it. The statistically-inferred climatology is interpreted in terms of current scientific understanding of the mechanisms through which related large-scale climatic variability affects the development and motion of Atlantic tropical cyclones. We also probe the spatial posterior probability distribution to quantify and rank the uncertainty in the climatology of TC landfalls that can be attributed to climatic variability, model parameter uncertainty, uncertainty in the historical landfall positions, a possible undercount bias early in the historical record, and sampling variability from the finite length of the observations. Given more detailed, expert information about uncertainty for each specific storm in the historical dataset, the model could be used to develop a definitive TC landfall climatology. It could also be used in conjunction with spatial information about exposures for risk management applications.