vertransform.f90 occasionally fails to initialize

[adingwell]

When running with smaller domains verttransform might fail to initialize ixm and jym.

At initialization the following block is run in verttransform.f90:

  ! Search for a point with high surface pressure (i.e. not above significant topography)
  ! Then, use this point to construct a reference z profile, to be used at all times
  !**************************************************************************************

    do jy=0,nymin1
      do ix=0,nxmin1
        if (ps(ix,jy,1,n).gt.100000.) then
          ixm=ix
          jym=jy
          goto 3
        endif
      end do
    end do
3   continue

In order to work, there has to be a grid point within the domain with a surface pressure above 1000 hPa. Such a point might not be present in smaller domains, where there is only a small number of grid points over sea surface.

It would be good if this block at least produces an error if no reference point is found.

[pesei]

I suggest that we take the lowest grid point found if there is none >1000 hPa.
I'll take the ticket for v9.2.

[adingwell]

I just noticed that this bug found its way into 10.2.

I've added the following lines between the loop and the continue statement in my local version:

ij_min  = MINLOC(ps(:,:,1,n))  ! ij-index of minimum pressure
ixm = ij_min(1)
jym = ij_min(2)

but I haven't tested it for a problematic domain yet (as my problems this time were related to #197)