《气象统计方法》-第七章 气象变量的正交分解.ppt

* First let me give a brief introduction to RFI The main characteristics of RFI include * * * * * The scattering plots reveal the relationship between 10 and 18 GHz over Asia. The red points satisfy this criteria, spectral difference larger than 5 degree kelvin We find that the spectral differences are negatively biased and linearly correlated. But over some regions TB at 10 can be much larger than TB at 18 GHz. * In Europe, the TB at 10 GHz are much larger than 340 degrees kelvin, which is the upper limit of the specified dynamic range of MWRI. It indicates that strong RFI is present in Europe. * Since the scattering effects from snow and ice are much stronger at high frequency than at lower-frequency channels, the differences between 10 and 18 GHz can also be more than 5 degrees kelvin over the regions with snow and ice. So besides this criteria, we add the spectral difference between 89 and 18 GHz and use the green color to indicate the very small and negative spectral difference. The green pattern over the large area in the north just shows the presence of snow, instead of RFI. * PCA data matrix is composed of the vector shown above. Then the covariance matrix S is constructed and its eigenvalues and eigenvectors are found satisfying this equation. The PC coefficients are obtained …… * Now let me give brief introduction to PCA. PCA is an efficient tool for analyzing the channel correlations. In our study we use PCA to separate RFI signals from natural radiations. The pca vector includes these five spectral indices. And NPCA use the normalized spectral indices. * The PCA detected RFI distributions are shown here. Left is for summer data and right is for winter. We see that without normalization, PCA can ……. The snow in high latitudes is mixed together with RFI signals. * Now let me give brief introduction to PCA. PCA is an efficient tool for analyzing the channel correlations. In our study we use PCA to separate RFI signals from natural radiations. The pca vec