Theses and Dissertations

Issuing Body

Mississippi State University

Advisor

Mercer, Andrew E.

Committee Member

Dyer, Jamie L.

Committee Member

Wood, Kimberly M.

Date of Degree

5-6-2017

Document Type

Graduate Thesis - Open Access

Major

Professional Meteorology

Degree Name

Master of Science

College

College of Arts and Sciences

Department

Department of Geosciences

Abstract

ropical cyclone-induced tornadoes (TCIT) exacerbate the devastation that landfalling tropical cyclones have on the United States. This research applied machine learning techniques in conjunction with midlatitude severe weather parameters to create an artificial intelligence (AI) capable of predicting TCIT occurrence. Severe weather diagnostic variables were collected at thousands of gridpoints from the North American Regional Reanalysis (NARR) to characterize the environments within tropical cyclones between 1991 and 2011. A support vector machine (SVM) was generated in various configurations to obtain the most effective AI. This approach revealed many parameters that were ineffective at predicting TCITs (primarily those utilizing the effective inflow layer). In addition, the most highly configured AI were capable of predicting TCIT occurrence with a Heidke Skill Score around 0.48.

URI

https://hdl.handle.net/11668/17558

Comments

statistics||artificial intelligence||tropics||weather||meteorology||forecasting||severe weather||tornado||hurricane||tropical cyclone

Share

COinS