Principal Component Analysis of Wind Profiler Observations

Abstract

Principal component analysis (PCA) is applied to wind profiler observations to study the vertical profile of the wind field and its temporal evolution. The rationale for decomposing time?height wind profiler data using PCA is twofold. The orthogonal vertical profile vectors are determined empirically from the variance of the observations, and the time evolutions of these vectors are not simple sinusoids, but are temporal varying signals that can be directly related to other measurements. As an example of its utility, PCA is used to compare the annual and interannual variation of zonal wind measured with a 50-MHz VHF wind profiler above Christmas Island, Kiribati, with the difference between surface pressures measured at Tahiti, French Polynesia, and Darwin, Australia. The high correlation coefficients relate the vertical profile of zonal wind observed in the central Pacific with the variation of convection in the western Pacific. Complex PCA (C-PCA) allows the analysis of data consisting of amplitude and phase information. It can describe the phase progression of oscillations embedded within the data. The C-PCA is applied to VHF wind profiler observations to study the seasonal behavior of the diurnal meridional wind observed above Biak, Indonesia, and the oscillatory structures of the vertical wind during a convective precipitation event observed above Darwin.