For calculating Prognostic Potential Predictability (PPP) on nostore-osl Mainly evaluating if the system is well constrained.

Author: Mao-Lin Shen (E-mail: maolin.shen@uib.no)

• Figures of the PPP for different levels are not shown because there are too many levels and GitHub can only present the web-page if the page size is smaller than 1MB.

• There are too many figures (over 32,000 figures and will be more). Therefore I have to seperate code and [the results (PPP-Results)] (https://github.com/TheEarnest/PPP-Results).

• Change the following head parameters in "cheeck_Prognostic_SNR"

  • DIR="the main output path"
  • tempDIR="where the temporary files are saved"
  • casePrefix="a general case name for all ensemble members"
  • MEyear="a year which must be existed in all ensemble members. For determining the file prefixes"
  • DataDIR="dir for the outputs"
  • ModelCompartments="ocn atm ice lnd"

• Before running the script (some known issues)

  • Make sure the data are well-prepared (no missing ensemble member or missing years)
  • micom outputs are compressed format, and cannot be processed by "cdo" because each file has different factor and offset. Better convert the outputs to F32 format before using this script.
  • The script of CAM automatically keeps the output of previous simulation, of which the exported variables are highly possible different with any new simulation. So those files (NorCPM_F19_tn21_mem??.cam2.h0.1979-12.nc in this case) should be removed in advance.
  • MICOM & ICE outputs have to be regridded before ploting
  • Main scripts are sitting @norstore-osl:/norstore_osl/projects/NS9207K/earnest/Analysis/Prognostic-Potential-Predictabil

• change all set of figures from PPP to SNR
• should add a script to check PPP evolving (comparing spatial std in this case, and focus on SST only) (Note: fldvar/fldcorr is not helpful)
• Should build a system to counter check the files
• uploading the nc files of PPP as well??
• From PPP to evaluate prediction barrier/uncertainty - comparing to ENSO prediction barrier
• add post process: skip some orginal output and adding more diag variables
• the ploting script has to be revised to get a better presentation, ie thiner colorbar, grids lines (not urgent)
• MICOM hy output is missing in the figure list (also fail plot the figures?)
• Trouble in presenting local maximum predictability!!! (HDF error).
• Should add the sctipt for calculating PPP with observation
• Should add a matlab script for ploting results of vertical section (mainly for MICOM)
• add a global map of SST bias or RMSE
• add Seasonal, 5-year pantad, decadal pantad results
• Sometime the model might write different number of variables when restart files were generated. (error handling is necessary)