Open funkpreacher opened 3 years ago
Same issue (standard interface). Looking at the source code, what seems to happen is every time you call "eofsAsCovariance", the original data is divided by the weights. So you can do it only once, or you must run the solver again... Other functions are not affected as they do not use the original data (only the PCs / EOFs), or weighting doesn't matter in the case of correlation.
Potential solutions would be
I am running the NAO example with xarray. Extracting the 1st EOF repeatedly gives different results (only when using weights). It looks like the weights are applied repeatedly.
Result of difference between eof1 and eof1b: