1. Eigen typhoon

Principal component analysis (PCA) is a method to calculate the direction of maximum variance in the feature space. In this project, we first classified typhoon images and created cloud-amount typhoon images with a reduced size. Then we regard an image array itself as a feature vector and calculated the eigenvectors of the whole typhoon image collection. An eigenvector is finally visualized as a gray-scale image, which we named an "eigen typhoon".

Northern Hemisphere

NWP ALL

Southern Hemisphere

SWP ALL

Especially Strong Typhoons in the Northern Hemisphere

NWP ALL (strong)

2. Multiresolution Analysis

The issue of "scale" is of paramount importance in meteorological phenomena. This is because any kinds of meteorological phenomena exist over only limited spatio-temporal scale. Here we focus on the issue of scale, and applied multiresolution analysis, or more specifically wavelet transform (A trous transform) to eigen typhoons.

1-st order eigen typhoon 1-st order eigen typhoon
30-th order eigen typhoon 30-th order eigen typhoon