Quote Originally Posted by CTA View Post
I'm getting exactly the results I want--as far as I can tell, every erroneous pixel is changed to the value of its nearest neighbor.

[...]

The only problem I encountered was that certain pixels were corrected to ocean in some maps and corrected to a prec/temp zone in others. To solve this, I wrote another short script that goes through all 4 images and converts those pixels to ocean.
That's pretty good output! The ocean vs. land problem shouldn't need a separate step once I've fixed up the ocean detection bug (the intended behavior of the climate script is already to only treat as land pixels that are land in all four input maps).