Implement SpAer-code used in CMIP6 #12
Comments
Kirkevag
added
enhancement
…d code for IRF-RFMIP)
|
The code has been tested on Vilje, but (with the correct settings for senith angle averaging and flux parameterisation) does not reproduce the (first month of) the original piClim-SpAer simulation for RFMIP, which has data available at A second identical test also gives different results. Similar tests on the (almost) original model version on Vilje /home/ntnu/kirkevag/noresm2.1dev have the same problem: A seemingly identical set-up as back then does not reproduce the first month, and neither does its twin case set up at the same time, also giving different results compared to the first twin. The differences are in meteorology and prognostic aerosols etc, but very small differences are found for the SpAer relevant prescribed parameters (such as anthropogenic AOD, AODVISSP, and multiplication gactor for CDNC, XCDNC_SP). The model code on /home/ntnu/kirkevag/noresm2.1dev is not identical to what it was in the CMIP6 simulatiion piClim-SpAer however. New code/scripts/compsets that were assumed not to be answer changing (and possibly were) has been pulled between July 30'th (the original code) and October 9'th and 24'th (after CMIP6 simulations). The following comments contain the logs from the two git pulls in October 2019. |
|
Pull log from October 9'th: -bash-4.2$ git pull La de eksperimentelle filene (inkludert for piClim-lu-oppsett) på error: Your local changes to the following files would be overwritten by merge: Endrer likedan der og prøver på nytt: -bash-4.2$ git pull |
|
Pull log from October 24'th: -bash-4.2$ git pull |
|
As a short summary, it can be said that the SpAer code, when active (#define SPAERO), does neither reproduce the CMIP6 piClim-SpAer2014 experiment, nor give the same results (meteorology-wise) for two identical cases set up in the new (updated) NorESM2 version. However, the prescribed SpAer optics and CDNC multiplication factors very close indeed to that in the above CMIP6 expriment, probably slightly different only because the meteorology differs. Not also that the code, when not active (#undef SPAERO) gives identical results in different experiments, as it should. Looking at the atm log-files, the difference in meteorology take place already in the 0'th time-step (nstep=0) (difference occur in the 7'th digit (for an example twin test): AM GLOBAL FIXER: 0.21274109505367922E-02 vs. AM GLOBAL FIXER: 0.21274178123676696E-02 My recommendations for further work with this code is to look for any intialized (newly introduced) variables, or erroneous coding with respect to pointers and physics buffers. |
|
The branch was technically working and gave comparable results to the original experiments in spring 2023. Should be added as a namelist option. |
|
This issue is now scheduled to be implemented in the NorESM2.3 release, as issue #103. |
The implementation of the SpAer code used in CMIP6 (e.g. for piClim-SpAer2014) has been put on hold to make sure that other CMIP6 simulations not could be affected by possible bugs in this code-wise quite big task.
The text was updated successfully, but these errors were encountered: