This is a page containing good practice tips for running WRF-LIS-CABLE as well as some known issues. It is a work in progress and will be updated as issues arise. Therefore, users are recommended to frequently check this page.

Good Practice

Vegetation Tiling

Most land surface models aim to capture land surface heterogeneity within the grid cell by implementing tiling where the grid cell is divided into tiles representing different plant functional types (PFT). At the point where the land surface fluxes are parsed to the atmospheric model, the flux estimates of each tile within the grid cell are aggregated to a grid cell estimate according to the PFT proportions of each respective grid cell (i.e. the PFT weights will vary by grid cell).

None of the templates implement vegetation tiling, this is intended for testing purposes. However you can modify this by updating the template_lis.config file. In particular, you can modify the number for the following option:

Maximum number of surface type tiles per grid:

You will also need to increase the runtime in the templates, as tiling will increase the computational costs.

Compilers and Software Versions

You will notice that in the templates there are particular software versions used that may not coincide with what you have previously used e.g. module load nco/4.7.7

PLEASE DO NOT CHANGE THESE! They have been set so that WRF-LIS-CABLE will run smoothly. If you have questions about why, please ask.

Parameter Datasets for LDT

For the LDT preprocessing tool there is a config file where you can stipulate the parameter datasets and their resolutions.

Most of these are available at a 1km resolution and therefore LDT can aggregate the parameter data to the resolution and projection of the model domain that you define.

However, this can take time! Therefore, we have already aggregated the parameter data to different global equidistant latlon resolutions of 10km, 25km, 50km, and 100km grids. This will speed things up.

It is best practice to use these regridded parameter datasets at the resolution closest to your domain i.e. if you are running a 12km resolution domain, use the 10km grids.

Computational Resources

All the templates have been setup to use an appropriate amount of CPUs, memory and walltime. However, if you make changes to the resolution, i/o frequency be mindful that you will need to run a quick test (e.g. a few timesteps) to understand whether you may need to make changes.

WRF Atmospheric Physics

Our WRF physics study Hirsch et al. 2019 can provide insights on optimal physics configurations for Australia only. It has not tested all possible WRF physics configurations. Further, it was applicable to the 50km resolution domain, it is likely that other configurations may be optimal for different configurations, particularly those that are convection permitting.

WRF suites are being developed and therefore it can be advantageous to check the WRF users site

Known Issues

• For CABLE there are parameters that LDT can process that we currently only have at 50km resolution global or 5km resolution for Australia only
• The albedo is lower than expected over central Australia
• If running with tiling it is possible when running LDT post-LIS to create the lis4real_input.d01.nc file that you have the odd grid cell with a very high soil temperature exceeding acceptable bounds of [170K to 400K]. This will cause real.exe to crash. If this happens, check your LIS.CABLE.YYYYMM-YYYYMM.d01.nc outputs and confirm that there is no unusual results. Then, to proceed to create the WRF boundary condition files you may use NCO to correct the soil temperature to a value close to the bound to enable the real.exe program to complete successfully. You could do this following the command:

  ncap2 -s 'where(SoilTemp > 400) SoilTemp=400' <path_to_bdy_data_subdirectory>/lis4real_input.d01.nc <path_to_bdy_data_subdirectory>/lis4real_input.d01.nc