NU-WRF Creation of the domain
Since LIS and WRF need to run on the exact same grid, the first thing to setup for NU-WRF is the grid and corresponding input datasets. The creation of the domain and the grid is relatively simple as it is the same as for WRF alone. To get an in-depth explanation please refer to the | WRF tutorial. The steps to follow are:
- Create a run directory for LIS anywhere you want. Then compile LIS only and copy the executable, LIS, from WRFV3/lis/make to the run directory (cp WRFV3/lis/make/LIS run_lis/.)
- Compile WRF and WPS. It's best to compile WRF with the lisreal option to enable the use of LIS else you'll need to recompile later.
- In WPS, update the namelist.wps to the domain you want. You may want to use util/plotgrids.ncl to help you design your grid.
- Run geogrid.exe
- Save the geo_em.dxx.nc files carefully as they are essential to share grid information between LIS and WRF.
Here are a few limitations and tips to keep in mind when designing your domain:
- You can not use a rotated lat-lon grid as LIS does not support those.
- Use the MAPFAC* variables output by WRF to ensure your grid is not too stretched in one or both directions. These should be close to 1 ideally. This is important for the code stability which it means being able to use larger time steps and hence to have faster simulations.
To run the testcase, make sure the namelist.wps file is the same as in the script directory NU-WRF_scripts/prepare_decks/. in step 3.
All input data available for LIS is stored under /g/data1/w35/LIS/LIS_PARAMS The global meteorological forcing data is under FORCING. It is advised to avoid using GDAS forcing over Australia. There is some MERRA forcing available for Australia under SUPP_FORCING. This dataset has been bias-corrected for the precipitation. All non-meteorological gridded inputs are under UMD. The highest resolution we have is 1 KM, more exactly it is on a lat-lon grid at 0.001° resolution (same naming convention for other resolutions). All the *_parms directories contain the LSM's specific input (e.g. soil and vegetation parameter files).
Interpolating the input dataset.
To speed up simulation time, one has to interpolate the non-meteorological fields at the required resolution beforehand rather than during simulation. However, LIS has basic interpolation built-in so it is not necessary to interpolate to the exact grid of the simulation. But it is better to interpolate the input fields to the same resolution as the simulation domain or at least at a close enough resolution. We are currently storing the input dataset in a few resolutions (including the 50KM resolution needed for the testcase) and we also provide the interpolation scripts to the users if they need to create new resolutions. If you think a new resolution of the dataset would be useful for several users, please contact [| climate_help] in order to add it to our collection.
Once you have cloned the repository with the scripts. Go into NU-WRF_scripts/RegridLIS. You will need to open and modify the regrid_all.csh file. You need to at least give the resolution (res), the output directory (outdir) and the input directories (indir and indir100 for PRINCETON dataset that is at 100KM). The script is setup to interpolate all datasets one by one. To speed up the interpolation process, you may want to comment out the datasets that won't be used by your LSM(s). Also, some interpolation programs accept command-line options which you might wish to change from the default. Refer to each interpolation program to know which options are available.