Redundant code in mo_edk_setvario.f90 file ?

[bibek-ux]

Hallo mHM/edk team,

I compiled the edk after cloning from git and ran it for regirding my precipitation data. However, the variogram plot plotted using variogram.dat produced unexpected results.(SS attached). The lag distance was way too much in each bin.

After some digging, I realized that the edk program seems to normalize, the lag distance before actual variogram fitting using the Nelder-Mead optimization method by dividing each lag distance in each bins by largest lag distance, however in the OPTI subroutine of mo_edk_setvario.f90 file, I found a redundant line of code that was basically, scaling the lag distance twice (1st and last line in the SS attached). I commented the redundant line , recompiled everything, changed the bin size and ran edk with openmp. I have also attached the empirical and variogram fitted plot after correction as well as the variogram values and confirmation stats.

I commented the second "where (gamma(:,1) > 0._dp) gamma(:,1) = gamma(:,1) * gmax(1)" assuming that "call stats" , requires lag distance to be in normalized state. Does this create any other problem for the code ?

Danke

[pallavhydro]

@StephanThober check this out :)

[MuellerSeb]

This is great you found that.
Thanks for digging into this.

I will fix this upstream when I find the time. Unfortunately we can't accept PRs on github since this is only a mirror!

Cheers, Sebastian

[bibek-ux]

Thank you Sebastian.