Extrapolation over complex terrain and propagation to real.exe

[stormchasegenie]

Colleagues,

As an experienced WRF user, I have an embarrassing question.

Build-up:
metgrid.exe provides the met_em files that contain pressure level data ingested by real.exe in the creation of the lower/lateral boundary condition, nudging, and initialization file.

In many driving isobaric datasets (e.g. ERA5), there are many below-ground gridcells that are populated with extrapolated values over complex terrain. If I set these same gridcells to something crazy (-9.99d+12) and run metgrid.exe, real.exe subsequently crashes. This indicates that real.exe may be pulling data from below-ground gridcells in the met_em files(?)

On a related note, in preprocessing GCM input binaries, I am finding that my wrf.exe results near areas of complex terrain are somewhat sensitive to whether or not I use vinth2p with numerical extrapolation or vinth2p_ecmwf with physical extrapolation (in NCL) at below-ground gridcells.

As a climate modeler, these differences disturb me.

Since my soil/skin temperatures do not suffer from this issue, and because this issue arises at a small percentage of my total gridcells, I am wondering just how much uncertainty is being introduced into my modeling framework as a result of changing the extrapolation method across gridcells that should be below-ground and thus not used(?)

This is a philosophical question, so I will not attach technical info for the moment...

-Stefan Rahimi, UCLA

 

[Ming Chen]

Stefan,
Please see my answers below:
(1) In many driving isobaric datasets (e.g. ERA5), there are many below-ground gridcells that are populated with extrapolated values over complex terrain. If I set these same gridcells to something crazy (-9.99d+12) and run metgrid.exe, real.exe subsequently crashes. This indicates that real.exe may be pulling data from below-ground gridcells in the met_em files(?)

For the output from metgrid, when the data is isobaric (as FNL probably is), the k=1 level is the surface, and k=2 through N are the isobaric surfaces (1000, 950, 900, etc hPa up through 1 hPa). It is not unusual in areas over topography that the surface level height (k=1) is larger than the k=2 1000 hPa height value (somewhere under the ground). This is not a concern. REAL program will handle this issue later.

Note that REAL doesn't accept missing values (i.e. crazy values in your case).

(2) On a related note, in preprocessing GCM input binaries, I am finding that my wrf.exe results near areas of complex terrain are somewhat sensitive to whether or not I use vinth2p with numerical extrapolation or vinth2p_ecmwf with physical extrapolation (in NCL) at below-ground gridcells.

My understanding is that vinth2p is commonly used for extrapolation, and it should work fine in WRF. Some can correct me if I am wrong.

(3) Since my soil/skin temperatures do not suffer from this issue, and because this issue arises at a small percentage of my total gridcells, I am wondering just how much uncertainty is being introduced into my modeling framework as a result of changing the extrapolation method across gridcells that should be below-ground and thus not used(?)

Honestly I don't know the answer. We ever really compare the impacts of the two different interpolation methods. Please keep us updated if you know something about this issue.