NCAR / VAPOR

VAPOR is the Visualization and Analysis Platform for Ocean, Atmosphere, and Solar Researchers
https://www.vapor.ucar.edu/
Other
176 stars 49 forks source link

CAM data sets: missing values not handled correctly #879

Open clyne opened 6 years ago

clyne commented 6 years ago

conversion of some CAM data sets to VDC not handled correctly.

To reproduce:

  1. Run: cfvdccreate Data/CAM/AdamPhillips/F2000C5_new120_025/FV_768x1152.bilinear.F2000C5_new120_025.cam.h0.2004-01.nc foo.nc
  2. Run: cf2vdc Data/CAM/AdamPhillips/F2000C5_new120_025/FV_768x1152.bilinear.F2000C5_new120_025.cam.h0.2004-01.nc foo.nc
  3. Run: vdccompare cf vdc Data/CAM/AdamPhillips/F2000C5_new120_025/FV_768x1152.bilinear.F2000C5_new120_025.cam.h0.2004-01.nc -- foo.nc

Note the large errors reported

clyne commented 5 years ago

Most of the variables look correct. However, there are excessive errors for the TROP* variables:

Testing variable TROP_P NLmax = 4.82252e+31 Testing variable TROP_T NLmax = 2.80003e+34 Testing variable TROP_Z NLmax = 8.41948e+31

This is due to the TROP variables not having missing values (even though missing_value attribute is defined) while other variables on same grid as TROP do have missing values. Bottom line: converting data sets to the VDC format is problematic if all variables defined on a particular grid do not have missing values defined at the exact same grid location.

Recommend postponing fixing this one