Hi Dave,
I have finally gotten back to testing this. I apologize for the delay. You're correct - that I previously was not using the wrfinput_d01 file you provided, as I was able to reproduce the problem without it. This round of testing, however, I used your wrfinput file, along with the most recent namelist you provided (I think - I am attaching it so that you can check it). The only thing I modified was to set frames_per_outfile = 1 so that when I diffwrf these output files, the output file for the initial run vs. the restart should contain the same times (just makes it easier for me). I ran a test with the namelist as-is (i.e., including RRTMG radiation), and then another test without radiation, but I still get diffs with both tests (with and without radiation).
To verify, I am currently using V3.9.1.1, compiled with Intel (V17.0.1.132), out-of-the-box code - no modifications, compiled with vortex-following, and for the tropical cyclone idealized case. I initially ran the case for 4 days (2015-09-01_00 to 2015-09-05_00), and then ran a restart for only 1 day, from 2015-09-03_00 to 2015-09-04_00, and then compared output at 2015-09-04_00:00:00.
Do you notice anything I'm doing differently than you?
I have finally gotten back to testing this. I apologize for the delay. You're correct - that I previously was not using the wrfinput_d01 file you provided, as I was able to reproduce the problem without it. This round of testing, however, I used your wrfinput file, along with the most recent namelist you provided (I think - I am attaching it so that you can check it). The only thing I modified was to set frames_per_outfile = 1 so that when I diffwrf these output files, the output file for the initial run vs. the restart should contain the same times (just makes it easier for me). I ran a test with the namelist as-is (i.e., including RRTMG radiation), and then another test without radiation, but I still get diffs with both tests (with and without radiation).
To verify, I am currently using V3.9.1.1, compiled with Intel (V17.0.1.132), out-of-the-box code - no modifications, compiled with vortex-following, and for the tropical cyclone idealized case. I initially ran the case for 4 days (2015-09-01_00 to 2015-09-05_00), and then ran a restart for only 1 day, from 2015-09-03_00 to 2015-09-04_00, and then compared output at 2015-09-04_00:00:00.
Do you notice anything I'm doing differently than you?