Testing the code for dynamic and physial variables and CFC11 as predictor and MLOTST, UVWMO, CFC12 as predictands with 10 sampling and 2 normalization schemes for the years 1960-1999. The out of sample predictions failed so far...

Testing the light-weight version of  the LSTM code on GPP (2 hours), as for 10 samples it takes too long, for CFC11 as predictor and MLOTST, UVWMO, CFC12 as predictands with 2 sampling and 2 normalization schemes

Testing the LSTM code for CFC11 as predictor and MLOTST, UVWMO, CFC12 as predictands with 10 sampling and 1 normalization schemes

The script to calculate the minmax maps for CFC A2ER_sixth v20200301 and UVWMO A5CW v20221230 absolute values on MN5 from two different experiments

The script to calculate the minmax maps for CFC A2ER_sixth v20200301 and UVWMO A5CW v20221230 absolute values on MN5 from two different experiments

Running now with 13 param 2 norms and 10 samples by adding UMO, VMO, WMO and checking how fast. The GPP account with only 8 CPU cores

The light-weight code version for MN5 with calculation of MaxMinMean and MaxMinMeanClim of Zet, Zeta 1 and Zeta 2, maps and series for Zeta1 and Zeta2 along with time consumption output. Runs for 1900-1999 within 3 hours for all 12 variables

The final script for the Min Max maps for 9 variables, histograms, and series plus the time output and plot output. Have to elucidate to suppress the warning text, as it occupies 500 M

Maps of MinMax Norms for all 9 variables. Cleaned version. Next to add timeseries for these extreme regions and their histograms of value distributions via argmax index

Experimenting with climatological min max norm anomaly maps to select the highest variation regions for training of the emulator

Full scale code with 9var 9sampl after testing the lightweight code 1var1samp for correction of RMSE, length of Forecast series plot and Heatmap scales of RMSE that are now up to 20%

LSTM module successfully run for INTPP and NO3. Test run completed. Next are  adding CORR_FORECAST and Heatmap for RMSE and CORR_FORECAST and GEO information for plots

Added geolocation info dictionary for index values and heat map. A bug seems to persist in ARIMA. Test now with 9VAR and 4Geoloc.