Michael's project blog

GFDL's bias

more cca

The results of yesterday's experiments suggest that weighting is not as effective as using more correlation EOFs. Of course, it's one experiment and there is lots of sampling error.

I'm going to re-do the calculations with all seasons and sets of predictors looking at up to 20 correlation EOFs. I'm also going to look at leaving 10 out (# of boots is 2*nt) instead of just one. Repeat with SST as predictor also.

western bias

Eli sent mail to GFDL asking about a warm bias that is seen in the western Pacific during the 1980's.

Re-sync

Need to get back to work and finish this project.

Mutiple predictors

How best to do CCA between gridded observations and model output with multiple predictors with disparate numbers of grid points?