when using match_geod_pergla, so far, the MB mean from 1999 to end of 2019 was calibrated to match the geodetic estimates. Normally it should be 2000-2019, but mb_climate_on_height falsely used the year before y0 even if hydro_month is equal to 1. This resulted in a not perfectly calibrated mb model and was found out during the development of this notebook: https://nbviewer.org/urls/cluster.klima.uni-bremen.de/~lschuster/error_analysis/working_glacier_gdirs_comparison.ipynb#id-dmdt-dmdtda-qc3. I proposed a correction for that. All old tests of OGGM still pass with this correction. If I am right and the corrections have to be done, we need to recalibrate the preprocessed gdirs above level 2 that use match_geod_pergla !

I also wrote a test that checks if the 20-year modelled specific mass-balance is now equal to the reference geodetic calibration data. Now it perfectly matches, however, and this is also tested inside of OGGM:

• For the test, we need the newest climate data (till end of 2019), however inside of https://cluster.klima.uni-bremen.de/~oggm/test_climate/, for both ERA5 and CRU, the climate goes only till 2018 or 2014,
  • Therefore, I use process_dummy_cru_file to create a climate in the right time range and to test then if the calibration has worked
  • I will create new pre-processed directories without the small error, for that I also added the climate_qc_months to the diagnostics if historical_qc_months is called!