Climate simulations typically extend for many months or years and into the future. For these simulations, reanalysis or future climate model data are best, as they provide a continuous data source.

Sea Surface Temperature (SST) data is extremely important in climate runs, so you will need a good source of SST, and to ensure that it is updated throughout the simulation period.

As this is only a test case, we are only running for 48 hours (and using past dates).

**Click on the links to jump to the desired section.**

Set-up and Run Geogrid
Set-up and Run Ungrib
Set-up and Run Metgrid
Set-up and Run Real & WRF

Set-up and Run Geogrid

• At this time, you may want to consider creating a new folder in which to save your met_em* files from previous runs, along with your namelist.wps file, in case you want to use them later and do not want to re-run everything from scratch.  


• For this climate simulation, you are creating a single domain over North America and the North Atlantic. Edit namelist.wps to reflect the following parameterizations:
  
  
  • The domain size will be 110 x 65 grid points.
  
  
  • Grid resolution will be 90 km.
    
    

    Note Downscaled climate simulations will typically use a finer resolution than 90 km, but since this is just a test case, and we would like to cover a large area, we are compromising on the resolution to save on computation and allow the simulation to run faster.

    
    
  
  
  • We will use a Mercator projection since our domain is closer to the equator.
  
  
  • To place the domain in the right location on the Earth, make the following additional changes to the namelist:

    &geogrid
    max_dom = 1,
    ref_lat = 15.00
    ref_lon = -60.00
    truelat1 = 0.0,
    truelat2 = 0.0,
    stand_lon = -60.00
    

Your domain should look like this:

 

Run geogrid.

Go back to top

Set-up and Run Ungrib

Degrib and reformat meteorological data from reanalysis or climate models. In this case, data come from NCEP GFS, available every 3 hours, and archived SST. We are going to ungrib the two datasets separately.

• First link the GFS data using the script link_grib.csh
   
  ./link_grib.csh   /glade/campaign/mmm/wmr/wrf_tutorial/input_data/climate/gfs.0p25.2016100600.f0
   
  
• Link the correct Vtable (data for this case are GFS, so use Vtable.GFS).



• Case dates are 2016-10-06_00 to 2016-10-08_00, and data are available every 6 hours.
  
   
• Run ungrib to create intermediate files (these files should have the prefix "FILE").



• Second, link the SST data using the script link_grib.csh
   
  ./link_grib.csh   /glade/campaign/mmm/wmr/wrf_tutorial/input_data/sst/rtg_sst_grb_hr_0.083.2016100
  
   
  
• Link the correct Vtable (input data are SST, so use Vtable.SST)
  
  
• Make the following changes to the &ungrib record of the namelist.

  &ungrib
  prefix = 'SST',

  Note The interval between SST data sets is 24 hours, but you need SST data 3-hourly to match the NCEP2 data interval. By keeping interval_seconds set to 21600, WPS will interpolate the available SST data to the required 3-hourly interval. Note the prefix change - this is so you do not overwrite the GFS data you just ungribbed. If the SST data had only been available, say weekly, we would have changed the namelist date range to the 2 available dates, before and after our case dates, and we would keep 'interval_seconds = 10800' to interpolate.


 
• Run ungrib to create intermediate files for SST data (the prefix of these files will be "SST").


 

Go back to top

Set-up and Run Metgrid

• Before running metgrid.exe, first make the following change to the &metgrid namelist record to ensure both FILE and SST files created by ungrib will be used by metgrid:

  &metgrid
  fg_name = 'FILE', 'SST'



• Now run metgrid, as usual.

Set-up and Run real and WRF

• Edit namelist.input to match the settings you used for WPS.
  
  

  Note Set "frames_per_outfile" to 1. This creates a new wrfout file for each output time, which is a good option for long runs, as it makes the output smaller and more manageable. For long runs, you will likely need to stop and start often, so create restart files (set restart_interval) at least once a day. This is not a restart run, though, so make sure you have restart = .false. Pay attention to the time_step. Since we are using 90 km grid spacing, this can be larger. The rule is that this must be set to a value of 6xDX or smaller. To ensure stability, set this to 360. Remember this is a single domain simulation. Be careful to not use too few vertical levels when doing long simulations, as this could lead to systematic biases. 45 levels should be okay for this case.

  
  
  
  
• Now add the following to the namelist.input file under the &time_control and &physics records.

  &time_control
  auxinput4_inname = "wrflowinp_d<domain>"
  auxinput4_interval = 360,
  io_form_auxinput4 = 2,
   
  &physics
  sst_update = 1,

  Note Adding the above lines ensures SST is updated every 6 hours throughout the model simulation. SST uses the auxillary stream 4, so we have to specify the file name that will be created when real.exe is run, the interval, and the output format (2=netcdf). Do not change the syntax for the auxinput4_inname parameter. WRF understand this syntax and will translate that to the correct file name.

  
  
  
• Now you are ready to run real and wrf. When running wrf.exe, use the runwrf_climate.sh batch script.

Go back to top

Organization Suggestion:

Recall the suggestion (from the "single domain" case) to create a directory to put your files in. Do this again for this case:

mkdir climate_run

and then move the necessary files into that directory to preserve them for potential later use.