An Overview and Evaluation of SynthETC: A Statistical Model for Extra-Tropical Cyclones

• Main Author: Uryayev, Rafael
• Format: Thesis
• Language: English
• Published: CUNY Academic Works 2019
• Subjects: Cyclones; Extra-tropical Cyclones; Statistical Modeling; Regression; Stochastic; Risk Assessment; Atmospheric Sciences; Climate; Meteorology; Probability; Statistical Models; Canada; Labrador Sea; North Atlantic
• Online Access: https://academicworks.cuny.edu/cc_etds_theses/768; https://academicworks.cuny.edu/cgi/viewcontent.cgi?article=1833&context=cc_etds_theses

Extratropical cyclones (ETCs) are the most common weather phenomena affecting the United States, Canada, and Europe. They can pose serious hazards over large swaths of area. In this thesis, a statistical model of ETCs, called SynthETC, is discussed. The model accounts for the for genesis, track path, termination, and intensity of statistically generated ETCs. Genesis is modeled as a Poisson process, whose mean is determined by climate and historical information. Tracks are modeled as a regression-mean determined by climate and historical information plus a stochastic component. Lysis is modeled using logistic regression, with climate states as covariates. Intensity is modeled using a resampling of historical intensities. A perturbation method is applied to the maximum intensities of all storms to allow the model to generate storms that are more intense than any in the historical record. Upon evaluation, two biases were identified: (1) not enough simulated storms moving northward and too many eastward in the region between the Labrador Sea and central North Atlantic, and (2) simulated storms move too fast with their distribution having a shorter tail than historical storm track steps. The model is evaluated under different modifications of its stochastic track component. This evaluation shows no significant improvement over the default, but does highlight the importance of the stochastic component of the track portion of the model. A potential pathway for model improvement would be to consider incorporating climate information into the deviation portion of the track component of the model.