Pattern forced geophysical vector field segmentation based on Clifford FFT

Vector field segmentation is gaining increasing importance in geophysics research. Existing vector field segmentation methods usually can only handle the statistical characteristics of the original data. It is hard to integrate the patterns forced by certain geophysical phenomena. In this paper, a template matching method is firstly constructed on the foundation of the Clifford Fourier Transformation (CFT). The geometric meanings of both inner and outer components can provide more attractive information about the similarities between original vector field and template data. A composed similarity field is constructed based on the coefficients fields. After that, a modified spatial consistency preserving K-Means cluster algorithm is proposed. This algorithm is applied to the similarity fields to extract the template forced spatial distribution pattern. The complete algorithm for the overall processing is given and the experiments of ENSO forced global ocean surface wind segmentation are configured to test our method. The results suggest that the pattern forced segmentation can extract more latent information that cannot be directly measured from the original data. And the spatial distribution of ENSO influence on the surface wind field is clearly given in the segmentation result. All the above suggest that the method we proposed provides powerful and new thoughts and tools for geophysical vector field data analysis.