[DISCUSSION] Stretched grid or nested grid for local mesh refinement

[geoschem]

[DISCUSSION] Stretched grid or nested grid for local mesh refinement

[JiaweiZhuang]

But, the normal cube sphere's vector bases are also different from the lat-lon's bases, so why is regridding onto the stretched grids vector bases any more challenging?

Because we shouldn't even regrid the lat-lon wind to cubed-sphere... The online GEOS simulation runs at cubed-sphere grid, but the output data are postprocessed/regridded to lat-lon grid -- this is how we get the GEOS-FP/MERRA2 archive for both GC-classic and GCHP. Our current model simulation is impacted by this double-regridding error which should have been avoided at all.

Ideally, the input wind data should be on the native cubed-sphere grid, and can be ingested by GCHP at the same resolution without regridding. Using lower-res grid or nested grid would just involve a simple averaging of fluxes at cell edges, but not rotating the vector bases. The stretched grid, on the other hand, requires more complex vector regridding (like the current lat-lon -> cubed-sphere) than just averaging fluxes.

[LiamBindle]

Hi @JiaweiZhuang,

Thanks again for writing this up. This has been really helpful.

When you have a change, could you expand on your concerns about regridding the wind fields and mass fluxes to a stretched grid? Are you concerned about regridding all vector fields in the met archives?

You mention that the cube sphere's vector bases could make regridding especially challenging. But, the normal cube sphere's vector bases are also different from the lat-lon's bases, so why is regridding onto the stretched grids vector bases any more challenging? I thought regridding between grids with different horizontal vector bases was routine. Doesn't ESMF's regridding support this?

Lastly, could you fill me in on the early wind and transport problems? I wasn't involved with GCHP at that point, so I'm not familiar with the issues.

cc: @lizziel, @sdeastham

Thanks,

Liam

[LiamBindle]

Ah I see, thanks, that clarifies things. Your concern is that the stretched grid could make what's an existing problem worse.

Why can't we get metfield archives on GEOS-5's native grid?

[JiaweiZhuang]

Your concern is that the stretched grid could make what's an existing problem worse.

Yeah, an exception is that, the online GEOS might also run at stretched grid and archive "native stretched grid" output. But this kind of "fixed stretched grid" seems too limited.

Why can't we get metfield archives on GEOS-5's native grid?

I guess asking for more archive from GEOS is a complex procedure.... Other people could better answer this.

[sdeastham]

It shouldn’t actually be hard; the problem is we need to talk to Bill Putman at GMAO about the process, and he hasn’t been responding to emails for the last few months. From: Jiawei Zhuang <[email protected]> Sent: Friday, September 13, 2019 1:32 PM To: geoschem/gchp <[email protected]> Cc: Sebastian Eastham <[email protected]>; Mention <[email protected]> Subject: Re: [geoschem/gchp] [DISCUSSION] Stretched grid or nested grid for local mesh refinement (#24) Your concern is that the stretched grid could make what's an existing problem worse. Yeah, an exception is that, the online GEOS might also run at stretched grid and archive "native stretched grid" output. But this kind of "fixed stretched grid" seems too limited. Why can't we get metfield archives on GEOS-5's native grid? I guess asking for more archive from GEOS is a complex process.... Other people could better answer this. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub<geoschem/GCHP#24>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ABSP43ARRXCG3LMEL4DKJXTQJPFB3ANCNFSM4HLFGRFQ>.

[LiamBindle]

I know this is a bit of a long shot, but is there any documentation of this double-regridding error? Or do you know if anyone has any simulation output that I could look at to get a better idea of how bad the problem is?

It shouldn’t actually be hard; the problem is we need to talk to Bill Putman at GMAO about the process, and he hasn’t been responding to emails for the last few months.

Maybe @lizziel and I should talk to them about getting access to these archives during our visit then. I'll add that to my list of things.

[sdeastham]

The archives don’t yet exist. The problem is that we need Bill to sign off on our proposed method to convert between dry and moist tracer mass mixing ratios: https://docs.google.com/document/d/1IQfDbb-kIyZSCTXxhaKyh8s1lP4woWszx_xp9edmCgQ/edit?usp=sharing The big question is whether we think more information is needed, and what the size of the errors will be from using interpolated surface pressure and specific humidity with time-averaged mass fluxes and courant numbers. From: Liam Bindle <[email protected]> Sent: Friday, September 13, 2019 1:46 PM To: geoschem/gchp <[email protected]> Cc: Sebastian Eastham <[email protected]>; Mention <[email protected]> Subject: Re: [geoschem/gchp] [DISCUSSION] Stretched grid or nested grid for local mesh refinement (#24) I know this is a bit of a long shot, but is there any documentation of this double-regridding error? Or do you know if anyone has any simulation output that I could look at to get a better idea of how bad the problem is? It shouldn’t actually be hard; the problem is we need to talk to Bill Putman at GMAO about the process, and he hasn’t been responding to emails for the last few months. Maybe @lizziel<https://github.com/lizziel> and I should talk to them about getting access to these archives then. I'll add that to my list. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub<geoschem/GCHP#24>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ABSP43HIHGJQVC7OZIL3ESTQJPGUBANCNFSM4HLFGRFQ>.

[JiaweiZhuang]

is there any documentation of this double-regridding error?

In early days (~2016) @sdeastham and I saw an excessive downward flux near the poles due to wind regridding. Here's the configuration I remember:

• Read input wind field at 4x5 lat-lon grid
• Run model at C48 resolution or higher
• Only turn on advection (disable other components) to highlight the error

This led to a lot of stratospheric ozone coming down to the surface. Haven't tried to reproduce this in recent versions. Not sure whether this issue is still there.

[LiamBindle]

Thanks, I'm going to take a closer look at this. I'll follow up early next week.

[lizziel]

This discussion thread has not been touched for a over year, although is still salient. @LiamBindle, would you like to transfer this to GCHPctm or close it out?

[LiamBindle]

@lizziel this can be closed