CCSM2.0.1 Known Bugs
- 1. The model does not restart exactly on Compaq machines. This problem is scheduled to be corrected in CCSM2.1 (reported May 17, 2002)
- 2. The ESMF library will not build with gmake versions 3.78.1 to 3.79 due to a recognized and corrected bug in gmake that involved file dating. This version of gmake seems to appear with IRIX 6.5. This problem can be corrected by updating gmake to version 3.79.1 or greater. (resolved June 15)
- 3. There is a error in hmix_gm.F in the ocean pop model. A 30-year test integration of the pop model with the corrected hmix_gm.F module resulted in small changes in the results: the average boundary layer depth changed by 2 - 2.5m in the convection regions in the GIN and Weddell Seas, the average heat flux into the ocean after 30 years was reduced by 0.09 W/m**2, changes in the barotropic streamfunction were less than 1Sv everywhere, and there were very small changes in the meridional overturning and poleward heat flux. A patch, patch1 is available on the CCSM2.0.1 download site to fix this error. The corrected version of hmix_gm.F is scheduled to be included in the CCSM2.1 release. (reported June 15, 2002)
- 4. The model does not thread correctly when running on a Compaq machine. This seems to be related to limitations in the operating system when running mulitiple hybrid parallel executables. This problem is being investigated further. (reported August 15, 2002)
- 5. There is a very minor bug in the coupler when passing tilt fields between the ocean and ice model. In the CCSM2.0.1 release, the tilt fields are being adjusted by an area correction in the coupler which should not be happening. This is corrected in patch 2 on the CCSM2.0.1 download site which provides a modification to the cpl5 file, ice_msg_mod.F90. This error changes the ocean tilt in the ice model by less than 0.1%. The corrected version of ice_msg_mod.F90 is scheduled to be included in the CCSM2.1 release. (reported August, 2002)
- 6. The latm.setup.csh script sets the latm namelist, DATA_DIR to read from the NCAR mass store system. This should be changed in all cases, even when running at NCAR, to copy from $CSMDATA. Users should edit latm.setup.csh, comment out the mss: statement and uncomment the cp: setting setting DATA_DIR (reported Oct 15, 2002).
#set DATA_DIR = mss:/CSM/DATA/ATM/T62/NCEP/4XDAILY
set DATA_DIR = cp:$CSMDATA/atm/latm5
- 7. CCSM will not restart exactly on seaborg.nersc.gov. This is due to a bug in the massvp3 library which was reported to IBM. There is a fix in patch 3 in the Macros.AIX file which sets the massv library to the default version. This patch results in exact restart and virtually no performance degradation. (reported October 10, 2002)
- 8. The CAM history output field Z050 (Z at 50 mb) is actually diagnosing Z005 (Z at 5 mb). This will be corrected in a CAM patch. (reported November 7, 2002)
- 9.CCSM does not compile correctly with the latest Fortran compilers on IBM machines with O3 optimization. O2 is now recommended. Patch 4 corrects this problem.
- 10.The component latm5 is sending the 4 short wave fields in the incorrect orde r. In msg.F90, the send buffer is set
buffs(i,j,16) = swvdr(i,j) ! sw down, near-infra, direct buffs(i,j,17) = swndr(i,j) ! sw down, visible direct buffs(i,j,18) = swvdf(i,j) ! sw down, near-infra, diffuse buffs(i,j,19) = swndf(i,j) ! sw down, visible duffuse
This should be
buffs(i,j,16) = swndr(i,j) ! sw down, near-infra, direct buffs(i,j,17) = swvdr(i,j) ! sw down, visible direct buffs(i,j,18) = swndf(i,j) ! sw down, near-infra, diffuse buffs(i,j,19) = swvdf(i,j) ! sw down, visible duffuse
This will impact the solutions. (reported Feb 5, 2003)
- 11. CCSM does not restart exactly in rare occasions when running with the active atmosphere model, CAM. This occurs if qneg4 triggers changes to some of the coupling fluxes on a restart timestep. This does impact answers slightly but should have no impact on climate. (reported June 10, 2003)
- Note: active CCSM bugs are reported on the ccsm-bugs tracking page.