Bug in bottom depth in MPAS-Ocean global_ocean init mode

[xylar]

A bug was introduced in #6310 in MPAS-Ocean's standalone init mode. In the global_ocean test case, the bottom depth is incorrectly altered, such that all but the shallowest layers have zero layer thickness.

As an example, see:
#6310 (comment)

This appears to be an incorrect change from a max() to a min() operator on this line:

E3SM/components/mpas-ocean/src/mode_init/mpas_ocn_init_global_ocean.F

Line 897 in ef925d1

bottomDepth(iCell) = min(bottomDepth(iCell), refBottomDepth(minimum_levels))

[xylar]

It isn't clear to me why 987126d, the commit that introduced this bug, was necessary. We compute minimum_levels above:

E3SM/components/mpas-ocean/src/mode_init/mpas_ocn_init_global_ocean.F

Lines 814 to 819 in ef925d1

	do k = 1, nVertLevels
	if (refBottomDepth(k).gt.config_global_ocean_minimum_depth) then
	minimum_levels = k
	exit
	end if
	end do

such that refBottomDepth(minimum_levels) is deeper than config_global_ocean_minimum_depth. It seems like the only circumstance where this would not be true is if config_global_ocean_minimum_depth is deeper than the deepest entry in refBottomDepth, and that should indicate that refBottomDepth should have been altered to accommodate the required config_global_ocean_minimum_depth.

Update: I think I may see. It's an issue when you have only one layer and refBottomDepth(1) == config_global_ocean_minimum_depth. Then, minimum_levels doesn't get set at all (which I think is still a problem).