Theses and Dissertations

Advisor

Mercer, Andrew

Committee Member

Dyer, Jamie

Committee Member

Rudzin, Johna

Date of Degree

8-7-2025

Original embargo terms

Immediate Worldwide Access

Document Type

Graduate Thesis - Open Access

Major

Professional Meteorology/Climatology

Degree Name

Master of Science (M.S.)

College

College of Arts and Sciences

Department

Department of Geosciences

Abstract

While Tropical Cyclone (TC) track forecasting has improved over the years it has been slow. These improvements can be attributed to increased computer capability and better model inputs. Machine learning has been used for forecasting different TC track and intensity. The study here looks to use a Random Forest to predict the TC latitude/longitude “landfall” at 72-hr and quantify improvements to the GFS forecast. Principal Component Analysis was used to reduce the dimensionality of the meteorological variables. Stepwise regression was used to determine the principal components that were important for forecasting landfall. These variables were then tested with a multivariate linear regression model as the control and then random forest in both training and testing sets of data. While there was not improvement to the GFS in the landfall position, there was improvement found with both methods tested here against another study that forecasted TC track error at 72 hours.

Included in

Meteorology Commons

Share

COinS