Principal Components

We begin by extracting the principal components for soil temperature. The intercept and slope is removed from each of the daily signals and the principal components are extracted. Eigenvectors corresponding to the second moment matrix for the different depths and different seasons for Soil temperature at Cub hill is shown in the figure above. The corresponding energies are also shown below the eigenfunctions.

Note that the eigenfunctions appear to be contaminated by outliers. A thing to try would be Tamas’s incremental and robust PCA to obtain a cleaner basis.