# How to Compute the EOFs by SVD?

• I

## Main Question or Discussion Point

Considering I have a matrix $\mathbf{A}$ which has a size of $M \times N$, how can I compute the Empirical Orthogonal Functions (EOFs) by Singular Value Decomposition (SVD)?

According to SVD, the matrix $\mathbf{A}$ is

$\mathbf{A} = \mathbf{U} \mathbf{\Sigma} \mathbf{V}^{T}$

where a superscript of $T$ denotes a transpose. Now, which are the EOFs in this equation, are they the rows of $\mathbf{V}^{T}$ or its columns (the rows of $\mathbf{V}$)?