bugs in the src/letkf_solver.F90

[xiaoyu20200515]

It seems that the code in Line 227 in the source code src/letkf_solver.F90:
obs_ij_rdiag(i) = obs_def(idx)%err
should be modified as follows:
obs_ij_rdiag(i) = obs_def(idx)%err**2

The reason is as follows:
By checking the codes we can found that the variable obs_ij_rdiag is used as the observation error variance in the subroutine letkf_core_solve in the source code src/letkf_core.F90.
However, the obs_def(idx)%err is read from the observation file, in which the variable err is given as the standard deviation of the observation error as decripted in the user-guide documentation.
Therefore, in my opinion, the variable obs_ij_rdiag should be the square of the variable obs_def(idx)%err.

[xiaoyu20200515]

I have made some land-surface assimilation experiments but got some unreasonble results and then found this bug. After fixing the bug, the experiment results looked better.