Nudging from CMIP6 input unrealistic geopotential

[epotter1]

Hello,

I am trying to run WRF using cmip6 data as input. I have been adapting GitHub - lzhenn/cmip6-to-wrfinterm: tools to process cmip6 data to drive wrf to create the files needed to input into metgrid.

My issue is, when interpolating the cmip6 data to pressure levels, I get unrealistic values below the surface. I did not think these would matter as they should not feed into the atmospheric data. However I am running with nudging, and getting very high values for PH_NDG (even high above the surface) and MU_BDG, which are somehow creating impossibly high precipitation. Can anyone help me with this? I am not sure why the below-surface interpolation is creating these problems into the atmosphere, or if there might be a different issue.

Many thanks,

Emily

 

[Ming Chen]

Hi Emily,
WRF doesn't allow unreasonable values in its intermediate and met_em files. This is because WRF terrain and terrain from large-scale forcing data could be quite different, and unreasonable input data may result in wrong information when running REAL program. This seems like the case you have seen.
Please handle under-ground data in a reasonable way.

 

[epotter1]

Hi Ming,

Thanks for your response. I will have another look at the interpolation, but I'm confused as to what's leading to this propagation up high in to the atmosphere. For example, looking at the GHT in the met_em.d01 file at 170 hPa, the geopotential varies slightly latitudinally, as expected:
Once I've run real.exe, in the equivalent wrfinput_d01 file at the same time, you can clearly see the topography at this level. This is the geopotential height from the wrfinput_d01 file at level (bottom_top_stag) 25 of 35. At this point, the model levels are almost exactly equivalent to pressure levels, with the pressure at model level (bottom_top) 25 equal to 171.0 hPa to 1 decimal place, and at model level (bottom_top) 24 equal to 197.6 hPa.

Although I think there is an issue with interpolation under the ground, this issue is specifically in the interpolation from the met_em file pressure levels to the WRF model levels.

I believe this is probably related to the values of pressure at the surface, which are used to create the WRF vertical coordinate. When I run using EAR5 data, the values of PSFC change slightly between the met file and the wrfinput file, from a minimum of 564 hPa to a minimum of 562 hPa. When I run using CMIP data, the values of PSFC change hugely between the met file and the wrfinput file, from a minimum of 621 hPa to a minimum of 773 hPa.

What is happening to the values of PSFC (and presumably other variables) during real.exe? And am I correct that if the PSFC values no longer correspond to the other data, it could be cause of the problem seem above?

Many thanks for your help,

Emily

 

[Ming Chen]

Emily,
I don't have an immediate answer to your question. This issue requires detailed examination of CMIP6 data. I suppose REAL can well handle input data from various model outputs and analysis products, and I am suspicious that your input data could be wrong. Can you check your met_em files and pay attention specifically to PSFC and PRES and under-ground data? It will be helpful to compare with other data like GFS and ERA5 to make sure the spatila pattern and vertical structure of CMIP6 data are correct.
Hope this is helpful.

 

[epotter1]

Hi Ming,

Thanks for this. it's CESM2 input data so must have been used to initialise WRF before, I think. In the met files, pressure is on constant levels which I've interpolated to (as I believe it has to be to run wrf?).

The initial pressure levels (from the raw CESM2 output) are derived from model levels using PSFC from the CESM2 data. However it seems like real.exe recalculates PSFC (presumably from PMSL). Is there a way to tell real.exe to use the input data from PSFC rather than recalculating? There seem to be some namelist options for this, but I'm not sure quite what they're doing. Are there any downsides to this approach (e.g. wrf model levels based on psfc no longer matching the wrf topographic height)?

I suspect the incorrect PSFC might be the problem as where the CESM2 data is from 'above ground' the WRF column is now using some data from under ground, however I vertically interpolate this will be incorrect, and it's quite a big depth under ground.