Should log(P) be interpolated linearly in height instead of P?

[jjguerrette]

The MPAS_to_physics subroutine interpolates pressure linearly with the height coordinate (see below).

MPAS-Model/src/core_atmosphere/physics/mpas_atmphys_interface.F

Line 397 in 7601162

pres2_p(i,k,j) = fzm_p(i,k,j)*pres_p(i,k,j) + fzp_p(i,k,j)*pres_p(i,k-1,j)

As the model levels get further apart near the top, this will tend to produce pressures that are too large since pressure varies approximately exponentially with height. It would be more accurate to interpolate log(P), then convert back to P afterward (similar to what is done for the model top extrapolation on L415). Although I do not know how pres2_p is used after this calculation, this could be a source of inaccuracy.

Note that the interpolation of P is inconsistent with core_init_atmosphere, which interpolates log(P).

See https://github.com/JCSDA-internal/mpas-jedi/issues/795 for more discussion.

EDIT: also see the related WRF publication, https://journals.ametsoc.org/view/journals/mwre/140/12/mwr-d-12-00045.1.xml

@liujake @mgduda @ldfowler58 @byoung-joo @metdyn

[1JunGu]

Like this request #1042 about isobaric height diagnostic using logarithmic interpolation

MPAS-Model/src/core_atmosphere/diagnostics/mpas_isobaric_diagnostics.F

Lines 613 to 614 in cd652c3

	! switch to use ln(pressure) for more accurate vertical interpolation, WCS 20230407
	pressure2(k,iCell) = exp(w1*log(pressure(k,iCell))+w2*log(pressure(k-1,iCell)))

The MPAS_to_physics and other similar subroutines should use this.
@mgduda @ldfowler58