Here are some results for application of EOF analysis to the surface cloud data. I used the same procedure as before (removal of seasonal cycle and cloud variability linearly related to Nino3.4 SST prior to the EOF analysis, and use of the correlation matrix), except that in this case I did not apply any rotation to the EOFs. The reason for this is that the surface cloud EOF patterns are very geographically widespread and therefore not suitable for rotation, which acts to emphasize local features.

I've attached the leading EOF pattern and time series for ocean-only low-level cloud amount and total cloud amount. The EOF time series show increasing trends that are even more linear than the global ocean-only averages, and the EOF spatial patterns resemble the global correlation patterns. The other EOF patterns and time series do not exhibit any large-scale organization or strange features in the time series, so I have not included them. I also have not included results for EOF analysis applied to land-only data since they also do not exhibit large-scale organization.