bug: mpas 2m temp, 10m wind interpolations

[hkershaw-brown]

💀 skeleton issue will fill in.

User reported 2M temperature observation, model_interpolate is using the 3D theta temp field

[hkershaw-brown]

similarly, for 2m (or 10m or whatever surface qtys) convert_vertical_state is using the 1st level of zgrid as the vertical location, should this be using 2m, 10m?

DART/models/mpas_atm/model_mod.f90

Lines 5012 to 5015 in e218864

	! surface obs should use the lower face of the first level. the rest
	! of the quantities should use the level centers.
	if ( ndim == 1 ) then
	zout(:) = zGridFace(1, cellid)

[nancycollins]

the first level of the 3d field is the model surface, and 2m and the surface are probably close enough. but the code does need to convert from theta (potential temperature) to sensible temperature - by computing with theta, rho, and qv. if it's not doing that with surface obs, then that's a bug.

if someone wants to interpolate 2m temps using the diagnostic 2m temperature field, i believe that is already sensible temperature and it shouldn't be corrected like the theta field is - just return the interpolated value.

[hkershaw-brown]

for completeness (coping from the email support - amazingly my skeleton issue is not enough information for this problem) user report that the 2m increments looks incorrect.
So maybe close enough but doing an (incorrect) conversion.
Or not close enough (with conversion correct).
Or maybe both not close and not correct.

to confirm:

1. conversion 'correct' theta (pot temp) -> 2m (temp) using level 1 of theta
2. surface obs (temp) -> using t2m (temp)
3. 1 is worse than 2.
4. t2m temperatures are quite different (cold bias?) output from MPAS

[criedel40]

This is such a tricky problem. If I am remembering correctly, the t2m output from MPAS is computed using one of the boundary layer physics packages in MPAS. Maybe I am remembering wrong? I am not sure if it is as simple as interpolation between say the first vertical model level above the surface and the surface (in MPAS). I think it might be more complicated. So while someone says "2m increments looks incorrect", could it be that the method for computing t2m in MPAS is not right? Maybe I am reading your email wrong. -Chris

…

On Mon, Nov 11, 2024 at 3:05 PM Helen Kershaw ***@***.***> wrote: for completeness (coping from the email support - amazingly my skeleton issue is not enough information for this problem) user report that the 2m increments looks incorrect. So maybe close enough but doing an (incorrect) conversion Or not close enough (with conversion correct) Or maybe both not close and not correct. to confirm: 1. conversion 'correct' theta (pot temp) -> 2m (temp) using level 1 of theta 2. surface obs (temp) -> using t2m (temp) 3. 1. is worse than 2. — Reply to this email directly, view it on GitHub <#768 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABIVLWCQBWBTIYEHYQLAYST2AEL3LAVCNFSM6AAAAABRJTJOCGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINRZGAYTSNZUGY> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[hkershaw-brown]

Thanks Chris, here is the clarification (apologies for doing a terrible job of reporting this issue!)

The 2m temperature is just much colder than the 3d temperature field (the plots I sent along earlier converted the 3d field from theta to temperature before plotting). Is there a known cold bias in the 2m temperature field for MPAS? It seems to have a really large cold bias for my fog cases.

[criedel40]

I mean, I wouldn't doubt that the t2m temperature field would have any bias. Like Nancy mentioned above, it is a diagnostic field so not predicted by MPAS. It could be sensitive to the choice of the PBL physics scheme and the land surface physics scheme in MPAS. I really don't think it is a simple interpolation between the first vertical layer above the surface and the surface temperature (skin temp or SSTs). An Interesting test/idea (maybe you already did this) would be to convert theta on the first model leve above the surface to temperature and using the surface temperature interpolate to 2 meters. Compare with the t2m in the output file. Maybe it would tell you something. -Chris

…

On Mon, Nov 11, 2024 at 3:23 PM Helen Kershaw ***@***.***> wrote: Thanks Chris, here is the clarification (apologies for doing a terrible job of reporting this issue!) The 2m temperature is just much colder than the 3d temperature field (the plots I sent along earlier converted the 3d field from theta to temperature before plotting). Is there a known cold bias in the 2m temperature field for MPAS? It seems to have a really large cold bias for my fog cases. — Reply to this email directly, view it on GitHub <#768 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABIVLWASAIXSTJE4RY5LJYL2AEN57AVCNFSM6AAAAABRJTJOCGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINRZGA2DGMJUGI> . You are receiving this because you commented.Message ID: ***@***.***>

[syha]

We should follow the approach used in wrf as mpas_atm adopted all the physics from wrf. For all surface DA, we want to use surface variables in the same way as in wrf/model_mod, where only horizontal interpolation is applied. In other words, we use 't2m' for 2-m temperature, with no vertical or variable conversion. As I had suggested, we should double check this for all surface obs and update the model_mod not to get into find_vert_indices. Please note that model error for the T2 bias is a separate issue. On Mon, Nov 11, 2024 at 2:29 PM Christopher Riedel ***@***.***> wrote:

…

I mean, I wouldn't doubt that the t2m temperature field would have any bias. Like Nancy mentioned above, it is a diagnostic field so not predicted by MPAS. It could be sensitive to the choice of the PBL physics scheme and the land surface physics scheme in MPAS. I really don't think it is a simple interpolation between the first vertical layer above the surface and the surface temperature (skin temp or SSTs). An Interesting test/idea (maybe you already did this) would be to convert theta on the first model leve above the surface to temperature and using the surface temperature interpolate to 2 meters. Compare with the t2m in the output file. Maybe it would tell you something. -Chris On Mon, Nov 11, 2024 at 3:23 PM Helen Kershaw ***@***.***> wrote: > Thanks Chris, here is the clarification (apologies for doing a terrible > job of reporting this issue!) > > The 2m temperature is just much colder than the 3d temperature field (the > plots I sent along earlier converted the 3d field from theta to temperature > before plotting). Is there a known cold bias in the 2m temperature field > for MPAS? It seems to have a really large cold bias for my fog cases. > > — > Reply to this email directly, view it on GitHub > <#768 (comment)>, or > unsubscribe > < https://github.com/notifications/unsubscribe-auth/ABIVLWASAIXSTJE4RY5LJYL2AEN57AVCNFSM6AAAAABRJTJOCGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINRZGA2DGMJUGI> > . > You are receiving this because you commented.Message ID: > ***@***.***> > — Reply to this email directly, view it on GitHub <#768 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA73GKPHUBXUA47U3H6RYPT2AEOVDAVCNFSM6AAAAABRJTJOCGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINRZGA2TCMJWG4> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[nancycollins]

someone pointed out to me that the bottom face of the first row of cells is the model surface, but theta is computed on cell centers, which are halfway up the thickness of the first level. sorry for adding to the confusion about surface obs.

[hkershaw-brown]

!MARINE_SFC_TEMPERATURE, QTY_2M_TEMPERATURE, COMMON_CODE
vs
!MARINE_SFC_TEMPERATURE, QTY_TEMPERATURE, COMMON_CODE
how general is this, e.g. cam-se, fv, ocean models, etc.
see Brett discussion pull request: #708 model_mod taking care of this. Switch out fwd operator depending on model? Obs def queries for QTY_2M first?