search

DanLi-BU / WRF

The official repository for the Weather Research and Forecasting (WRF) model
Other
1 stars 0 forks source link

How to "borrow" (unused) NUDAPT variables in SLUCM for other purposes? #6

Open sunt05 opened 1 year ago

sunt05 commented 1 year ago

The following is drafted by @DanLi-BU with light edits by @sunt05

OK, now at least I think I can elaborate my questions clearly. Here they are:

  1. in wrfinput_d04 (attached), there is a variable called BUILD_HEIGHT, which corresponds to HGT_URB2D in the WRF code according to this line in Registry.EM_COMMON:

https://github.com/DanLi-BU/WRF/blob/3dd1546ba425b5e77d682dd92253738416b3cb0a/Registry/Registry.EM_COMMON#L813

  1. HGT_URB2D/BUILD_HEIGHT is one of the key variables in NUDAPT. If HGT_URB2D is larger than 0, it will trigger using NUDAPT, see https://ral.ucar.edu/sites/default/files/public/product-tool/NUDAPT_44_Documentation.pdf This is also corroborated by looking at the function urban_var_init in module_sf_urban.F

  2. Here is the question. If I manually modified BUILD_HEIGHT in wrfinput_d04, say make it 10 m everywhere, and save it as BUILD_HEIGHT in wrfinput_d04. This has NO effect on my WRF results. Well, now that I think about it more, I start to understand why, because WRF actually never has a variable called BUILD_HEIGHT (or never reads in a variable called BUILD_HEIGHT from wrfinput_d0x). The variable name in the code is HGT_URB2D, even though BUILD_HEIGHT is the variable name in wrfinput_d04.

  3. Similarly (but also not exactly), there is a variable called STDH_URB2D in wrfinput_d04 (again see attached), which is also called STDH_URB2D in the WRF code.

https://github.com/DanLi-BU/WRF/blob/3dd1546ba425b5e77d682dd92253738416b3cb0a/Registry/Registry.EM_COMMON#L815

But when I modified it, there was again no effect on the WRF outputs.

  1. Then I searched STDH_URB2D in all .F files and here is what I got:
./WRF//phys/module_sf_urban.F:2832:                             HGT_URB2D,MH_URB2D,STDH_URB2D,                & ! inout
./WRF//phys/module_sf_urban.F:2928:   REAL, DIMENSION( ims:ime, jms:jme ), INTENT(INOUT) :: STDH_URB2D
./WRF//phys/module_sf_urban.F:2976:               IF (HGT_URB2D(I,J)>0. .or. STDH_URB2D(I,J)<0.) THEN
./WRF//phys/module_sf_urban.F:2981:                  WRITE(mesg,*) 'STDH_URB2D',STDH_URB2D(I,J)
./WRF//phys/module_sf_urban.F:2988:                     STDH_URB2D(I,J)=0.
./WRF//phys/module_sf_urban.F:3013:               IF (HGT_URB2D(I,J)>0. .or. STDH_URB2D(I,J)<0. ) THEN
./WRF//phys/module_sf_urban.F:3018:                  WRITE(mesg,*) 'STDH_URB2D',STDH_URB2D(I,J)
./WRF//phys/module_sf_urban.F:3025:                     STDH_URB2D(I,J)=0.
./WRF//phys/module_sf_urban.F:3050:              IF (HGT_URB2D(I,J)>0. .or. STDH_URB2D(I,J)<0. ) THEN
./WRF//phys/module_sf_urban.F:3055:                  WRITE(mesg,*) 'STDH_URB2D',STDH_URB2D(I,J)
./WRF//phys/module_sf_urban.F:3062:                     STDH_URB2D(I,J)=0.
./WRF//phys/module_sf_urban.F:3087:              IF (HGT_URB2D(I,J)>0. .or. STDH_URB2D(I,J)<0. ) THEN
./WRF//phys/module_sf_urban.F:3092:                  WRITE(mesg,*) 'STDH_URB2D',STDH_URB2D(I,J)
./WRF//phys/module_sf_urban.F:3099:                     STDH_URB2D(I,J)=0.
./WRF//phys/module_sf_urban.F:3130:                   STDH_URB2D(I,J)=0.
./WRF//phys/module_physics_init.F:156:                         HGT_URB2D,MH_URB2D,STDH_URB2D,          & !Optional multi-layer urban
./WRF//phys/module_physics_init.F:666:   REAL, OPTIONAL, DIMENSION( ims:ime, jms:jme ), INTENT(INOUT) :: STDH_URB2D !SLUCM
./WRF//phys/module_physics_init.F:1419:                HGT_URB2D,MH_URB2D,STDH_URB2D,                  & !Optional multi-layer urban
./WRF//phys/module_physics_init.F:2429:                HGT_URB2D,MH_URB2D,STDH_URB2D,                  & !Optional multi-layer urban
./WRF//phys/module_physics_init.F:2778:    REAL, OPTIONAL, DIMENSION( ims:ime, jms:jme ), INTENT(INOUT) :: STDH_URB2D !SLUCM
./WRF//phys/module_physics_init.F:3137:                              HGT_URB2D,MH_URB2D,STDH_URB2D,                   & !urban
./WRF//phys/module_physics_init.F:3254:                              HGT_URB2D,MH_URB2D,STDH_URB2D,                   & !urban
./WRF//phys/module_sf_noahmpdrv.F:2938:    STDH_URB = STDH_URB2D(I,J)
./WRF//dyn_em/module_initialize_real.F:3023:               grid%STDH_URB2D(i,j)  = grid%URB_PARAM(i,93,j)
./WRF//dyn_em/module_first_rk_step_part1.F:707:     &        ,MH_URB2D=grid%mh_urb2d,STDH_URB2D=grid%stdh_urb2d          & !SLUCM
./WRF//dyn_em/start_em.F:1137:                      grid%MH_URB2D,grid%STDH_URB2D,grid%LF_URB2D,                     & !SLUCM

  1. My understanding is that STDH_URB2D is created after run real.exe (the line of code highlighted earlier, also see here, but there was never a place where the WRF model "reads in" the variable STDH_URB2D from wrfinput_d04. So it makes sense that modifying it does not affect the simulation results.

  2. But then when there is indeed valid NUDAPT data (as we did in previous weeks), how does real.exe "pass" STDH_URB2D, as well as HGT_URB2D, to wrf.exe through wrfinput_d0x? Does this have anything to do with this grid%? Interesting...

sunt05 commented 1 year ago

The interesting exploration by @DanLi-BU does reminds me of the coupling work we did a while ago for WRF-SUEWS.

  1. But then when there is indeed valid NUDAPT data (as we did in previous weeks), how does real.exe "pass" STDH_URB2D, as well as HGT_URB2D, to wrf.exe through wrfinput_d0x? Does this have anything to do with this grid%? Interesting...

And it's right - the custom variables in wrfinput are indeed loaded via the grid%xx derived type - see here

So we will need to tweak the related part in module_first_rk_step_part1.F to get this working (hopefully...).

DanLi-BU commented 1 year ago

This is the modified wrfinput_d04. I changed STDH_URB2D but there are no effects.

wrfinput_d04_mod.nc.zip

DanLi-BU commented 1 year ago

This is the namelist.

namelist.input.zip

sunt05 commented 1 year ago

Please upload all input files so I can run this case at my side.

DanLi-BU commented 1 year ago

namelist

namelist.input.zip

original wrfinput

wrfinput_d01.zip wrfinput_d02.zip wrfinput_d03.zip wrfinput_d04.zip

modified wrfinput

wrfinput_d01_mod.nc.zip wrfinput_d02_mod.nc.zip wrfinput_d03_mod.nc.zip wrfinput_d04_mod.nc.zip

wrfbdy

wrfbdy_d01 1.zip.001.zip wrfbdy_d01 1.zip.002.zip wrfbdy_d01 1.zip.003.zip wrfbdy_d01 1.zip.004.zip wrfbdy_d01 1.zip.005.zip wrfbdy_d01 1.zip.006.zip

DanLi-BU commented 1 year ago

wrfinput_d02.zip

DanLi-BU commented 1 year ago

wrfinput_d03.zip

DanLi-BU commented 11 months ago

This is the code I used to modify the wrfinputs

wrfinput_d04_modification.zip

sunt05 commented 11 months ago

_Now we've resolved an issue with the STDH_URB2D variable, which lost its metadata during a Python-based modification process - just wanted to write down the interesting debugging journey that might help fellow WRF users/coders._

🔍 Initial Search & Dead Ends

💡 Breakthrough with Compilation Files

🛠️ Debugging Steps Taken

Implemented some debug lines using the existing code structure:

WRITE ( message , FMT = '("processing wrfinput file (stream 0) for domain ",I8)' ) grid%id
CALL wrf_debug              ( 100 , message )

Remember to adjust the debug_level in your namelist.input!

Added checks right after reading the NetCDF file revealed unexpected all-zero values for grid%stdh_urb2d.

🕵️‍♂️ Comparing Variables Checked another urban-related variable, frc_urb2d, which loaded correctly – indicating an isolated issue with just one variable.

✅ Solution Identified A few lines of quick xarray code pinpointed that while modifying values, we didn't preserve attributes (attrs). The fix? Modify only the necessary variable values rather than replacing entire objects.

sunt05 commented 11 months ago

This is the code I used to modify the wrfinputs

wrfinput_d04_modification.zip

Below is the core part of the zipped ipynb:

Click to expand! This is the content that will be hidden until the section is expanded. You can include text, images, code, etc. ```python import xarray as xr import pandas as pd import matplotlib.pyplot as plt import numpy as np # List of file names to modify file_names = ['d01', 'd02', 'd03', 'd04'] # Base path for the files base_path = "wrfinput_" # Iterate over each file, modify, and save for file_name in file_names: # Open the dataset ds_path = base_path + file_name ds = xr.open_dataset(ds_path) # Your existing code for modifying stdh_urb2d and hgt_urb2d stdh_urb2d = ds['STDH_URB2D'] hgt_urb2d = ds['MH_URB2D'] frc_urb2d = ds['FRC_URB2D'] lu_index = ds['LU_INDEX'] # Define the condition for LU_INDEX lu_condition = lu_index.isin([23, 24, 25, 26]) # Update values based on the conditions stdh_urb2d1 = xr.where((frc_urb2d > 0) & lu_condition, -4.5, stdh_urb2d) hgt_urb2d = xr.where((frc_urb2d > 0) & lu_condition, 15, hgt_urb2d) # Assign the modified variables back to the dataset ds['STDH_URB2D'] = stdh_urb2d1 ds['MH_URB2D'] = hgt_urb2d # Save the modified dataset to a new file modified_ds_path = base_path + file_name + "_mod.nc" ds.to_netcdf(modified_ds_path) print(f"Saved modified dataset to {modified_ds_path}") ```

Updated version:

Click to expand! ```python # List of file names to modify file_names = ["d01", "d02", "d03", "d04"] # Base path for the files base_path = "wrfinput_" # Iterate over each file, modify, and save for file_name in file_names: # Open the dataset ds_path = base_path + file_name ds = xr.open_dataset(ds_path) # Auxiliary variables frc_urb2d = ds["FRC_URB2D"] lu_index = ds["LU_INDEX"] # Define the dictionary of variables with their new values dict_var_mod = { "STDH_URB2D": -4.5, "MH_URB2D": 15, } # Define the condition for LU_INDEX lu_condition = lu_index.isin([23, 24, 25, 26]) # Update values based on the conditions for var, val in dict_var_mod.items(): ds[var] = ds[var].where((frc_urb2d > 0) & lu_condition, val) # Save the modified dataset to a new file modified_ds_path = base_path + file_name + "_mod.nc" ds.to_netcdf(modified_ds_path) print(f"Saved modified dataset to {modified_ds_path}") ```