Hi Glenn - I agree with you that getting an incorrect plot with no indication of error is the worst fear. What if it got published? The only solution is to do a calculation several different ways, which is ordinarily the case anyway as you seek to understand a result. With respect to your example, I have no definitiev answer, but just a suggestion: rather than work with the entire grid, work with a single latitude (perhaps writing out two new netcdf files at the same latitude). Often the details behind a phenomena become clearer when plotted with gridpoint dots (plot/line/sym=27) than when trying to compare color-fill plots by eye. The fact that GrADS is also making incorrect plots would make me look very closely at the data in those files, rather than at the software ..... Good luck. Billy K