Open eguil opened 7 years ago
@eguil glad this has been caught.. Have you figured out the source of the error yet?
no but I suspect vertical interpolation issues in weakly stratified regions. We will need to enquire. I wrote an ad hoc corrector (correctBinFiles.py) that I now run as a pre-processor to the mme calculations. These are things we should have looked at in more details earlier... but Yona's work is helping do this now !
Paul Barker has a code to remove vertical instabilities and that we could use in a preprocessing stage when we re-run everything.
@eguil there are also some vertical interpolation (spline) routines that were planned to be incorporated in the matlab version of the TEOS-10 library.. This should also be considered..
See #42
Bug corrected today for bottom interpolation. The density range was larger than specified in s_s and hence some polar or warm pool profiles did not fin any density bin. Check if this help the pb above: it did not. Debugging continuing...
Bug still present. Should check if new library of ESMF called 'esmpy' solves the problem.
@eguil the latest cdms2
is now 3.1.2. This comes with ESMF/ESMPy 7.1.0r, the latest release - see https://anaconda.org/conda-forge/cdms2/:
cdms2 conda-forge/osx-64::cdms2-3.1.2-py27hcfef303_5
cdtime conda-forge/osx-64::cdtime-3.1.2-py27h3edfc39_0
...
esmf conda-forge/osx-64::esmf-7.1.0r-1
esmpy conda-forge/osx-64::esmpy-7.1.0r-py27h24bf2e0_2
ptopsoxy and other ptop xy variables wrong (S=0 !) in some places. See attached screen shot:
May be due to unstable columns where interpolation or test fails as density profile is not constinuously increasing with depth.
Way out is to test if ptopsoxy = 0 and apply properties at surface from 3D fields (i.e. first non-masked density level). Can be done as a post-processing but should also be corrected in binDensity.py