Issue with Negative Surface Pressure in WRF-ARW Simulation Using ERA5 and ERA5-Land Data

[Chetan]

Hi Everyone,

I’m currently simulating the WRF-ARW model using a combination of ERA5 and ERA5-land data. Here’s the setup I’m using:
- For most surface variables, I use ERA5-land data.

- If ERA5-land data is unavailable, I use ERA5 surface layer data.

- All pressure level variables come from ERA5.

This approach works well when the simulation domain is entirely over land. However, when the domain extends into sea regions, I encounter an issue where the met_em files display negative surface pressure values. This causes errors in real.exe. The problem seems to arise because ERA5-land data is only available over land and contains NaN values over the sea, which leads to discrepancies. In contrast, using ERA5 data for both surface and pressure level variables does not produce this error.

Has anyone experienced a similar issue or worked with a similar setup? How did you resolve it? Any suggestions or advice would be greatly appreciated.

I’ve attached the namelist.wps file for reference.

Thank you in advance!

Best,
Chetan

 

[Ming Chen]

Chetan,
I am suspicious that the issue is caused by wrong interpolation of SST ( and probably other variables such as soil variables.)
Can you check your met-em files and see whether there are missing values of SST, SM and ST ?

 

[Chetan]

Thanks for your response, @Ming Chen.

Most grid points in the met_em file show zero SST, but a few points have negative values in PSFC, SM1, and ST1 (first layer).

The error real.exe gives is "Could not find level above ground."

I followed the user guide to use data from multiple sources. I thought NaN values for surface variables in ERA5-land would be replaced by values from ERA5, but it doesn’t seem to be working that way.

Also, I resized (lower size) the domain for ERA5-land so that it covers only the land area. I expected surface variables to come from ERA5-land wherever it overlaps the bigger domain and those for other nonoverlap areas to come from ERA5. However, metgrid.exe automatically resizes the bigger domain created for ERA5 to the smaller ERA5-land domain.

 

[Ming Chen]

Hi Chetan,
The input data definitely is problematic. Please see details below.

Thanks for your response, @Ming Chen.

Most grid points in the met_em file show zero SST, but a few points have negative values in PSFC, SM1, and ST1 (first layer).

PSFC, SM1 and ST1 shouldn't be negative. If they are negative, please check the data you ungrib, --- it is possible that these data are wrong

The error real.exe gives is "Could not find level above ground."

I followed the user guide to use data from multiple sources. I thought NaN values for surface variables in ERA5-land would be replaced by values from ERA5, but it doesn’t seem to be working that way.

WPS cannot process NaN data.

Also, I resized (lower size) the domain for ERA5-land so that it covers only the land area. I expected surface variables to come from ERA5-land wherever it overlaps the bigger domain and those for other nonoverlap areas to come from ERA5. However, metgrid.exe automatically resizes the bigger domain created for ERA5 to the smaller ERA5-land domain.

 

[Chetan]

WPS cannot process NaN data.

So, let's say ERA5-land data contains the NaN values (ex PSFC variable) for one corner grid point; then, does the metgrid.exe take the PSFC from ERA5 data for this grid point?

I have set the metgrid section as follows:
&metgrid
fg_name = 'ERA5', 'ERA5_Land'

I have attached the images for my domain D01, which touches the Sea area at the top right corner, where ERA5-land contains the NaN values and the reanalysis data for ERA5 and ERA5-land.

 

[Ming Chen]

if the same variables are available in both ERA5 and ERA5-Land files, then the data from ERA5-Land will be used.
In your case, if any variable in ERA5-Land has NaN value, it will be processed by metgrid.exe and cause trouble later.

 

[Chetan]

Thank you, @Ming Chen, for the clarification.