A
Comparison of the Karhunen-Loeve (KL) and Maximum Noise Fraction (MNF)
Algorithms on Axis-Symmetric Hurricane Model Dataa
Amanda Fox, Michael Kirby, Michael Montgomery, and John Persing
The Karhunen-Loeve (KL) and Maximum Noise Fraction (MNF) algorithms attempt to derive eigenvectors and eigenvalues to reduce the complexity and redundancy of the dataset. These techniques are often used in image and signal processing but have not been applied to hurricane data. The objective of this research is to compare the Karhunen-Loeve (KL), also known as Empirical Orthogonal Function (EOF) or Principal Components Analysis (PCA), and Maximum Noise Fraction (MNF) algorithms on axis-symmetric hurricane model data. Eight days of five-minute output data is derived from an axis-symmetric hurricane model that simulates flow velocity in the vertical, radial and azimuthal directions, temperature, pressure, and liquid water and water vapor content. Results will attempt to determine the dimension of the hurricane and its primary modes of variation.