A potential bug in the 'u_baro' calculation of FALWA v.1.2.0

[sandrolubis]

Hi all, after updating to FALWA v1.2.0, I spotted a potential bug in u_baro calculation where the output mirrors - i.e., SH = NH. I checked the 'interpolated_u' data; they are correct. Only 'u_baro' shows this issue. See the attached plot.

[chpolste]

I can reproduce the problem with QGFieldNHN22. QGFieldNH18 does not seem to be affected.

I think the loop in

hn2016_falwa/falwa/f90_modules/compute_flux_dirinv.f90

Lines 190 to 193 in e47f685

	do j = jstart,jend ! ### yet to be multiplied by cosine
	ubaro(:,j) = ubaro(:,j)+uu(:,j+nd-1,k)*exp(-zk/h)*dc
	urefbaro(j) = urefbaro(j)+uref(j-jb,k)*exp(-zk/h)*dc
	enddo

needs to be adapted to the value of is_nhem as in the loop earlier in the function. Maybe remove

hn2016_falwa/falwa/f90_modules/compute_flux_dirinv.f90

Lines 181 to 188 in e47f685

	! Bounds of y-indices in N/SHem
	if (is_nhem) then ! 5N and higher latitude
	jstart = jb+1 ! 6
	jend = nd
	else
	jstart = 1
	jend = nd-jb ! nd - 5
	endif

and replace everything with

    if (is_nhem) then
      do j = jstart,jend
        ubaro(:,j) = ubaro(:,j)+uu(:,j+nd-1,k)*exp(-zk/h)*dc
        urefbaro(j) = urefbaro(j)+uref(j-jb,k)*exp(-zk/h)*dc
      enddo
    else
      do j = jstart,jend
        ubaro(:,j) = ubaro(:,j)+uu(:,j,k)*exp(-zk/h)*dc
        urefbaro(j) = urefbaro(j)+uref(j,k)*exp(-zk/h)*dc
      enddo
    endif

@csyhuang, do you think this works with the indices? It seems to resolve the issue in my test notebook, but I'm worried about an off-by-one error...

[csyhuang]

@sandrolubis Thanks for reporting the bug!
@chpolste Thanks for figuring out the cause of this issue! I think your solution works (maybe also define explicitly jstart and jend in the loop?)! Would you like to submit a bugfix PR (and update version number to v1.2.1)? If not, I can do it this weekend after work.
Thanks again 🙏 😊