Inverted EOF correlation on ERSST example

[rgvill1217]

Hi,

The EOF applied to SST in this notebook seems to have its correlation signs inverted (https://github.com/pangeo-data/pangeo-ocean-examples/blob/master/noaa_ersst_variability.ipynb)

While sst_anom_detrented is plotting reasonable anomalies (this corresponds to the Mega-El Niño of 97-98):

and while ONI indexes are quite reasonable too:

The EOF's leading mode is not behaving like the aforementioned, plotting a sign-inverted canonical El Niño mode

PC1 is also sign-inverted. The last two are not consistent with Eof's function developer example (https://ajdawson.github.io/eofs/latest/examples/elnino_standard.html)

Perhaps it could be related to the weighting? I can not figure out why Eof function is not plotting EOF and its PC the way it should be.